This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 1 Losses of soil organic carbon with deforestation in mangroves of Madagascar 2 Shortened version for page headings: Losses of soil carbon with mangrove deforestation 3 Ariane Arias-Ortiz1,2*, Pere Masque1,3,4,5, Leah Glass6, Lisa Benson6,7, Hilary Kennedy8, Carlos M. 4 Duarte9, Jordi Garcia-Orellana1,4, Claudia R. Benitez-Nelson10, Marc S. Humphries11, Ismaël 5 Ratefinjanahary6, Jaona Ravelonjatovo6, Catherine E. Lovelock12 6 1 Institut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona, Bellaterra, 08193 Barcelona, Spain. 7 2 Ecosystem Science Division, Department of Environmental Science, Policy and Management, University of California, 8 Berkeley, CA, USA. 9 3 School of Science & Centre for Marine Ecosystems Research, Edith Cowan University, Joondalup WA 6027, Australia. 10 4 Departament de Física, Universitat Autònoma de Barcelona, Bellaterra, 08193 Barcelona, Spain. 11 5 International Atomic Energy, 4a Quai Antoine 1er, 98000, Principality of Monaco, Monaco. 12 6 Blue Ventures Conservation, Villa Bella Fiharena, Rue Gambetta, Lot 269, Toliara 601, Madagascar. 13 7 Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 OHT, UK. 14 8 Ocean Sciences, Bangor University, Menai Bridge, Anglesey LL59 5AB, UK. 15 9 King Abdullah University of Science and Technology (KAUST), Red Sea Research Center (RSRC) & Computational 16 Bioscience Research Center (CBRC), Thuwal, 23955-6900, Saudi Arabia. 17 10 School of Earth, Ocean & Environment, University of South Carolina, Columbia, South Carolina 29208, USA. 18 11 Molecular Sciences Institute, School of Chemistry, University of the Witwatersrand, Johannesburg, South Africa. 19 12 School of Biological Sciences, The University of Queensland, St Lucia, QLD 4072, Australia. 20 21 *correspondence to aariasortiz@berkeley.edu; 22 Manuscript highlights: 23 • Deforestation led to the loss of 20% of 1-m C stocks and increased soil homogeneity 24 • Annual C loss in deforested soils was 4.5 times the C sequestration of intact soils 25 • Conservation is temporally more effective than restoration for reducing C emissions 26 27 28 1 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 29 Abstract: Global mangrove deforestation has resulted in substantial CO2 emissions to the 30 atmosphere, but the extent of emissions from soil organic carbon (C) loss remains difficult to assess. 31 Here, we sampled 5 intact and 5 deforested mangrove plots from Tsimipaika Bay, Madagascar, to 32 examine the loss of soil C in the 10 years since deforestation. We estimated tree biomass and 33 analysed grain size, 210Pb activities, organic C and total nitrogen (N) and their stable isotopes in soils 34 as well as dissolved organic C in surface waters. Deforested soils revealed evidence of disturbance in 35 the upper 14 g cm-2 (∼ 40 cm) when compared to reference intact soils, indicated by lower porosity, 36 higher dry bulk density, an order of magnitude higher soil mixing and loss of C and N despite no 37 significant soil erosion. While C loss from biomass was unequivocal and was estimated at 130 Mg C 38 ha-1, the C loss from soils was more difficult to assess given the large heterogeneity of intact forest 39 soils. We estimated that the loss of C due to mangrove clearing and soil exposure over 10 years was 40 equivalent to ~ 20% of the upper meter soil C stock, and ~ 45% of the C stock accumulated during 41 the last century. Soil C loss rate was 4.5 times higher than the C sequestration rate in reference intact 42 soils. These results emphasize the importance of mangrove conservation for CO2 emissions 43 mitigation, as they suggest that deforestation-C losses will take substantially longer to offset with 44 mangrove restoration. 45 46 Keywords: Mangroves, deforestation, soil carbon, soil disturbance, CO2 emissions, Madagascar. 47 48 2 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 49 1. Introduction 50 Mangroves are an important natural resource in the tropics and sub-tropics and provide a 51 wide range of ecosystem services, including coastal protection, support of fisheries and biodiversity, 52 nutrient cycling and carbon sequestration (Barbier and others 2011). Their high rates of primary 53 production and the low rates of organic matter decomposition in their flooded soils lead to 54 mangroves having some of the highest soil organic carbon (C) stocks among forested ecosystems 55 (Donato and others 2011; McLeod and others 2011). However, the C stored in their soils is 56 vulnerable, as mangroves have been widely degraded and converted to alternative land-uses, 57 resulting in losses of ecosystem services (Alongi 2002) and significant carbon dioxide emissions 58 (CO2) to the atmosphere as previously stored C is remineralized to CO2 (Lovelock and others 59 2017b). 60 Consequences of mangrove soil disturbance include subsidence; as roots die, soil volume 61 collapses and erosion occurs, resulting in loss of soil C. These effects are generally most severe and 62 longer lasting when caused by anthropogenic versus natural perturbations (Ellison and Farnsworth 63 1996; Twilley and Day 2012). Soil C loss has been observed, for instance, where soils have been 64 excavated for the construction of aquaculture ponds (Ong 1993; Kauffman and others 2014). At sites 65 where mangroves have been removed or uprooted due to human activities or where there have been 66 intense storms, losses of soil C have been inferred from changes in soil elevation (Cahoon and others 67 2003; Lang’at and others 2014) or measured as CO2 efflux (Lovelock and others 2011; Sidik and 68 Lovelock 2013; Lang’at and others 2014). Although deforestation has been the major cause of forest 69 loss in the past, there are few studies that have directly measured the change in soil C content in 70 mangrove soils when forests are degraded, but soils remain in place (Kauffman and others 2016; 71 Grellier and others 2017; Adame and others 2018). 3 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 72 A change in soil C content with mangrove deforestation may occur directly as a consequence 73 of biomass loss and/or indirectly due to factors affecting soil biogeochemical processes, such as a 74 change in temperature, physical protection or aeration. For example, variations in mangrove biomass 75 may reduce inputs of labile C and nutrients from detritus, and increased soil temperature from direct 76 sun exposure may result in further loss of soil C (Granek and Ruttenberg 2008). Destructive 77 practices, such as clear cutting, mechanically redistribute C in the soil (Yanai and others 2003; 78 Zummo and Friedland 2011; Lundquist and others 2014), enhancing oxygen diffusion, altering 79 microbial communities and organic C remineralization (Kristensen and Alongi 2006). Physical 80 mixing processes modify the soil structure, transport conditions and soil chemistry, hence potentially 81 changing the availability of C and N for associated microbial and plant communities (Balesdent and 82 others 2000). For instance, the addition of biodegradable C to deep stable C stores promotes 83 microbial respiration of the added and existing organic pools through a priming mechanism (Bianchi 84 2011), and may accelerate C mineralization beyond that directly derived from mechanical mixing, 85 contributing to elevated CO2 emissions from soils. Soil C may also be lost as a consequence of soil 86 erosion or exported as dissolved C, a process which may be accelerated through direct exposure of 87 mangrove soils to tidal inundation, rainfall, and waves (Thampanya and others 2006; Labrière and 88 others 2015), as well as through changes in the composition and biomass of benthic mats driven by a 89 reduction in mangrove litter loading (Delgado and others 1991; Duke and Wolanski 2001; McKee 90 2011; Grellier and others 2017). 91 92 Overall, soil C storage represents a balance of C inputs and losses, and because soil is the 93 largest reservoir of C in many mangrove ecosystems (Donato and others 2011), small changes in its 94 pool size may translate into significant CO2 emissions and changes of C fluxes to coastal waters 95 (Atwood and others 2017; Gillis and others 2017). While it is well established that the harvesting of 96 terrestrial forests results in a loss of C in organic and mineral soil horizons (Yanai and others 2003; 4 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 97 Diochon and others 2009; Zummo and Friedland 2011), the impact of mangrove deforestation on soil 98 C storage remains limited. Direct assessment of these C losses is important for quantifying CO2 99 emissions from mangrove deforestation and in the value associated with avoiding emissions achieved 100 through conservation projects (Herr and others 2017). Currently, guidance by the International Panel 101 for Climate Change on CO2 emission from coastal wetlands (IPCC 2014) has a “tier 1” (i.e., default) 102 assumption that soil CO2 emissions and removals are zero for forest management practices in 103 mangroves. However, the IPCC allows for country-specific “tier 2” (i.e., based on direct 104 assessments) to be adopted by using a C stock-difference method in order to account for any 105 emissions associated with forest management practices. 106 In this study we aim to quantify the C loss from deforested mangrove soils in northwest 107 Madagascar. Madagascar contains Africa’s fourth largest national extent of mangroves, representing 108 approximately 2% of the global mangrove area. Since 1990, more than 20% of Madagascar’s 109 mangrove ecosystems have been heavily deforested because of the increasing demand for charcoal 110 and timber by urban populations (Jones and others 2016a), which is the primary cause of mangrove 111 deforestation in the broader East African region (FAO [Food and Agriculture Organization of the 112 United Nations] 2007). We sampled soil cores from plots that were either within deforested and 113 intact forest areas to assess the change of soil C after clearing. We described the physical 114 characteristics of soils with depth, quantified the variation in C and N contents, and estimated 115 sediment accumulation rates, soil mixing and potential erosion using the natural radionuclide lead- 116 210 (210Pb) (T1/2: 22.3 yr) (Appleby 2001; Arias-Ortiz and others 2018). We also measured 117 dissolved organic C (DOC) in adjacent coastal surface waters to assess the effects of deforestation on 118 DOC export 10 y later. Finally, we use our data to estimate the fate and total change of soil C since 119 mangrove clearance. 120 121 2. Methods 5 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 122 2.1 Study site 123 This study was conducted in Tsimipaika Bay (previously referred to as Ambanja Bay; Jones 124 and others 2016a) in northwest Madagascar (48°28’E, 13°30’S), where anthropogenic mangrove loss 125 is particularly prominent due to extensive extraction for charcoal and timber (Jones and others 2014). 126 Together with Ambaro Bay, the Tsimipaika-Ambaro Bay complex forms Madagascar's second most 127 extensive mangrove ecosystem, with over 40,000 ha of mangrove forests (Jones and others 2014, 128 2016a). The site is characterized by a humid sub-tropical climate and is influenced by semi-diurnal 129 tidal ranges varying between maximums of 3.0 – 3.5 m (Rasolofo and Ramilijaona 2009). Mangrove 130 soils in this region are underlain by alluvial and lake deposits (Jones and others 2016b) flooded by 131 sea level rise. Contemporary localized mapping supported by field observations indicated that, in 132 2010, anthropogenic activities had driven substantial deforestation (1,000 ha) mostly near the 133 southwest region of the peninsula that separates the two bays (Jones and others 2014) (Fig. 1). 134 Deforestation heavily targets closed canopy mangrove forests followed by open canopy mangroves 135 (Benson and others 2017), which represent 30 and 56% of the total mangrove area at the Tsimipaika- 136 Ambaro Bay complex, respectively (Jones and others 2014). Rhizophora mucronata is favored for 137 the charcoal production process, in which trees are felled by hand and carried to nearby temporary 138 kilns to be carbonized. Non-Rhizophora species and unwanted prop roots are burned to heat kilns, 139 although large volumes of downed wood are often discarded at the site. 140 141 In November 2016, we sampled soil cores from 5 plots that were cleared between 2006 and 142 2008 and from 5 plots from an intact forest (Fig. 1). The deforested and forested plots were spatially 143 separated by ∼5 km but had relatively similar environmental conditions such as species composition, 144 hydroperiod, type and proximity of bedrock and nutrient inputs. Within the sampled plots, four 145 species of mangroves were present: Rhizophora mucronata, Bruguiera gymnorhiza, Ceriops tagal 146 and Sonneratia alba (Table 1). Intact plots were all well-formed, closed (> 60%) canopy mangroves, 6 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 147 consisting of high stature trees (mean height: 9.0 ± 0.5 m) of variable density (800 - 4,700 ha-1). The 148 average diameter at breast height (1.3 m, dbh) was 12 ± 2 cm. Deforested plots used to contain 149 closed canopy mangrove forests (Blue Ventures, personal communication) and this was evident from 150 plots comprised of Ceriops tagal or Bruguiera gymnorhiza where the boles left in place after tree 151 removal lead to stump densities ranging from 1,000 to 11,400 ha-1 (Table 1). 152 153 Aboveground tree biomass at each plot (Table S1) was derived from tree diameter and height 154 measurements using generalized species-specific allometric equations (Clough and Scott 1989; Cole 155 and others 1999; Chave and others 2005; Kauffman and Donato 2012) and wood density values 156 (Chave and others 2009; Zanne and others 2009). These equations were chosen based on the region 157 in which they were developed and the parameters used to derive them, and have been previously 158 described in Jones and others (2016b) for use in the same area. The maximum tree diameter and 159 height measured (18 cm and 10 m, respectively) were within the range of trees used to develop the 160 equations. Tree belowground-biomass was calculated using the generalized equation presented in 161 Komiyama and others (2005) and equations from Kauffman and Donato (2012) were used to 162 estimate the biomass of standing dead wood. From biomass density estimates, the total biomass C 163 stocks (Mg C ha-1) were estimated using conversion factors of 0.50 and 0.39 for above and 164 belowground estimates, respectively (Kauffman and Donato 2012), and should therefore be 165 considered as approximate estimates of biomass C. Plot size for tree measurements was 100 m2 and 166 estimates of biomass and C stocks were scaled to the hectare-level. 167 168 2.2 Sampling and analytical methods 169 In order to characterize soil biogeochemical properties, PVC tubes (1.5 m long, 6.2 cm inner 170 diameter) were hammered into mangrove soils (one at each plot; Fig 1), extracted by hand and 171 transported to the laboratory. Prior to extracting the core, the depth to the sediment surface inside and 7 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 172 outside the core was measured in order to assess core shortening during sampling (Glew and others 173 2001), which averaged 36 ± 10% and 50 ± 13% in intact and deforested mangrove soils, 174 respectively. The PVC corers were cut lengthwise, and the soils inside the corers were sliced at 0.5 175 cm-thick intervals throughout the first 20 cm, and at 1 cm-thick intervals below this depth. Soil depth 176 layers were weighed wet and then dried at 60ºC until a constant weight was achieved. Soil water 177 content and dry bulk density (DBD) were then calculated. Soil mass per unit area (g cm-2) was also 178 estimated at each layer by dividing the dry sample mass by the core tube area sampled. The lack of a 179 reference elevation marker in sampled intact and deforested soils made comparisons of soil 180 properties based upon depth or volume limited and less precise (Gifford and Roderick 2003; Wendt 181 and Hauser 2013). We thus used the cumulative mass approach. The cumulative soil mass of each 182 core was calculated by summing their respective soil mass layers to the bottom of the core. Soil 183 profiles were displayed in terms of cumulative mass rather than depth to accommodate the impact of 184 changes in the DBD, and thus surface elevation change, which may occur through soil collapse after 185 deforestation (Cahoon and others 2003; Krauss and others 2010; Lang’at and others 2014), the 186 influence of trampling (Kauffman and others 2004), swelling and shrinking with changes in moisture 187 content (Haines 1923) and due to variable soil shortening and compaction during coring. 188 Soil C and total nitrogen (subsequently notated as N) contents were measured at 1 cm 189 resolution throughout the upper 30 cm, and in alternate slices every 5 cm below this depth. Prior to 190 analysis, soil samples were sieved (1.5 mm) to exclude belowground biomass before being ground to 191 a fine powder. Sub-samples (~ 20 mg) were weighed into silver cups, acidified with 1 M HCl, dried 192 at 60ºC and then analysed using an elemental analyser (Carlo Erba NA1500). No visual evidence of 193 effervescence was observed during sample acidification, which indicates that minimal inorganic C 194 was present. Analytical precision (s.d. of n = 26) was ± 0.3% for C, ± 0.02% for N and ± 4 for molar 195 C:N ratios. Stable isotopes of sediment C and N (13C and 15N) were analyzed at one core per 196 treatment location using an elemental analyzer–isotope ratio mass spectrometer (Hilo Analytical 8 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 197 Laboratory) at the University of Hawaii. Sub-samples for stable isotope analyses were encapsulated 198 in silver cups and acidified as described above. The use of a weak 1 M HCl solution was chosen 199 following the recommendations in Kennedy and others (2005). Replicate and control samples (NIST 200 8704) were also run and the accuracy and precision of 13C and 15N data were of ± 0.2‰ and ± 201 0.07‰, respectively. 202 Grain size analyses were conducted down core to evaluate potential erosion, which results in 203 selective and preferential loss of smaller size grain fractions (Arata and others 2016). Sediment 204 grain-size was measured with a Mastersizer 2000 laser diffraction particle analyzer following 205 digestion of bulk samples with hydrogen peroxide to remove organic matter. Sediments were 206 classified as sand (63 - 1,000 m), silt (4 - 63 m) and clay (< 4 m) (size scale: Wentworth 1922). 207 Specific activities of 210Pb were measured down core in order to assess soil accumulation 208 rates and soil erosion. Total 210Pb was determined through the analysis of its granddaughter 210Po by 209 alpha spectrometry after complete sample digestion in a HNO3:HF mixture (9:3 ml) using an 210 analytical microwave in the presence of a known amount of 209Po added as a tracer (Sanchez-Cabeza 211 and others 1998). Certified reference materials IAEA-447 and IAEA-385 were analyzed alongside 212 soil samples. Accuracy of the 210Pb(210Po) measurements averaged 96 ± 4%. The specific activities 213 of 210Pbxs used to obtain the age models were determined as the difference between total 210Pb and 226 214 Ra (supported 210Pb). Specific activities of 226Ra were determined for selected samples within each 215 core by gamma-spectrometry through the measurement of 226Ra decay product emission lines of 214 216 Pb at 295 and 352 keV and using calibrated geometries in a HPGe detector (CANBERRA, Mod. 217 SAGe Well). Total 210Pb activities at depth derived by alpha and 226Ra specific activities via gamma 218 were within error of one another confirming agreement between alpha and gamma methods. Mean 219 sediment accumulation rates over the last several decades to century were estimated for intact 220 mangrove soils using the Constant Flux:Constant Sedimentation (CF:CS) model applied below the 221 surface mixed layer (Krishnaswamy and others 1971) following the recommendations in Arias-Ortiz 9 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 222 and others (2018). Mass accumulation rates (MAR) are expressed in cumulative dry mass units (g 223 cm-2 y-1) and accretion rates (SAR) in mm y-1. Carbon accumulation rates (CAR) were estimated as 224 the product of the fraction of %C accumulated down to the excess 210Pb horizon (Ct) by the MAR of 225 that period (MARt): 226 𝐶𝐴𝑅 = 𝐶𝑡 · 𝑀𝐴𝑅𝑡 (Eq.1) 227 The same equation can be applied to estimate N accumulation rates. Potential soil erosion triggered 228 by mangrove removal was assessed by comparing the 210Pbxs inventories (210Pbxs activity per unit 229 area) at deforested mangrove soils against the inventories measured in intact mangrove soils. 230 Soil C stocks were quantified using equivalent soil mass layers rather than depth to account 231 for variations in bulk density across sites and down core. The soil mass layers used as the basis for 232 comparison were 14 g cm-2 (upper), which represented soils accumulated in the last century (based 233 on the average 210Pbxs horizon), 14 - 45 g cm-2 (bottom), and 45 g cm-2 (total), which was the average 234 cumulative mass equivalent to approximately 1 m of soil. Estimates of soil C stocks integrated to a 235 depth of 1 m are also reported for comparison to global estimates. For the latter purpose, we 236 corrected soil depth for core shortening by linearly distributing the spatial discordance between the 237 length of the recovered soil and the depth penetrated by the core tube to the sliced soil layers 238 following Morton and White (1997). We acknowledge that by using a whole-core value, differential 239 interval shortening with depth might be underestimated and could lead to erroneous conclusions 240 regarding the magnitude of deformation and the need for depth corrections. Here, cumulative mass 241 intervals were used as a reference for detailed depth-dependent quantitative comparisons. A different 242 coring method that reduces shortening (Hargis and Twilley 1994) or the taking of several intermittent 243 measurements of soil shortening during coring (Morton and White 1997) would have been necessary 244 to use depth-based approaches. 245 Surface water samples were collected over a tidal cycle at 8 sites along the shore adjacent to 246 the intact (stations 1-4) and deforested (stations 5-8) mangrove areas. The intact sites were sampled 10 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 247 during the flood to high slack tide. The deforested sites were sampled during the ebb to low slack 248 tide. Samples for total organic C were collected in 20 mL combusted glass vials, acidified to pH of 2 249 and stored at 4°C until analysis. Samples were analyzed as both filtered (using acid-washed and pre- 250 combusted syringe GF/F filters) and unfiltered samples. Total and dissolved organic C were 251 measured using high temperature (720°C) catalytic oxidation (Pt-alumina) on a Shimadzu TOC-V 252 CPN analyzer (Benner and Strom 1993). Analytical replication (5 injections, 100 μL) of consensus 253 reference material (Florida Straight at 700 m, DOC-CRM program) was run every 10 samples. High 254 concentrations were also measured diluted with distilled and deionized water. Total organic C versus 255 filtered DOC concentrations, and high concentration versus diluted samples were found to be the 256 same within error (± 50 μmol C L-1). 257 258 2.3 Statistics 259 Water content, DBD, C and N content and molar C:N ratios in soils as well as DOC 260 concentrations in surface waters were not normally distributed, and thus non-parametric tests were 261 used to assess significant differences between intact and deforested areas (Mann-Whitney test) at a 262 level of significance of < 0.05. We used principal component analysis (PCA) to assess the 263 relationship between soil properties and the two areas studied (intact and deforested mangroves). 264 Sediment grain size, 210Pbxs inventories, 13C and 15N, and stocks did follow a normal distribution, 265 hence a two-sample t-test was used to assess significant differences between intact and deforested 266 mangrove soils. Mean ± SE values are reported throughout the manuscript together with median 267 values where variables were not normally distributed. 268 269 2.4 Emissions from mangrove forest deforestation 270 We analysed C losses and potential emissions from mangrove deforestation based upon the 271 variation in C content in the mangrove soil profiles between intact and deforested soils. Likewise, 11 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 272 biomass C loss was estimated as the difference between vegetation C stocks in intact and deforested 273 plots. The losses of C were reported as CO2 equivalents (CO2e), obtained by multiplying C loss 274 values by 3.67, i.e., the molecular ratio of CO2 to C. The mean annual rate of C loss from deforested 275 soils was estimated as the total soil C stock loss divided by the time elapsed since disturbance (10 276 years). 277 278 3. Results 279 The effect of deforestation was obvious due to differences in mangrove vegetation compared 280 to undisturbed sites. Total estimated biomass C stocks in intact plots ranged from 113 to 254 Mg C 281 ha-1, while in deforested plots these were substantially lower (0.06 to 9.5 Mg C ha-1) and contained 282 negligible belowground biomass (i.e., roots) (Table 1 and Table S1). The effects of deforestation 283 were less clear in soils from deforested mangroves, which contained similar average C and N 284 contents to intact mangroves over the total soil profile (P > 0.05) (Table 2). This was in part because 285 variability in soil properties within the five intact mangrove cores was large and data followed a 286 bimodal distribution (Fig. S1). Intact soils Cc19 and Cc20, hereafter referred to as high-DBD intact 287 soils, were depleted in water content and had significantly higher soil DBD and lower C and N 288 contents relative to deforested soils (P < 0.01). In contrast, intact plots Cc18, Cc28 and Cc29, 289 hereafter referred to as low-DBD intact soils, had significantly lower DBD and higher water, C and 290 N contents than deforested soils (P < 0.01). As a consequence, physico-chemical properties of 291 deforested soils fell between those of high-DBD and low-DBD intact soils (Table 2). 292 Multivariate analyses further confirmed that soils from intact mangroves were represented by 293 two clusters of data and that deforested soils were comparable to a mixture of low- and high-DBD 294 intact soils, with characteristics closer to low-DBD intact soils (Fig. 2). Principal components Pc1 295 and Pc2 explained 75% of the total variance among sampled soils. Pc1 comprised 50% of the 296 variance and was strongly correlated with soil DBD (r = -0.90), water, and C and N contents (r = 12 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 297 0.95, r = 0.95 and r = 0.94, respectively). Pc2 explained 25% of the variance and was strongly 298 correlated with clay (r = 0.91) and moderately correlated with C:N ratios (r = 0.53). 299 300 Concentrations of DOC in surface ocean waters varied widely between stations, with 8 times 301 higher median concentrations measured during ebb tide (Fig. 3). Stations 1- 3, adjacent to the intact 302 mangrove area and sampled during the flood tide, had the lowest DOC concentrations (320 ± 30 303 mol C L-1) contrasting with those measured at station 4 (3,500 ± 50 mol C L-1), also adjacent to 304 the intact forest, but sampled during the high slack to ebb tide. All surface waters nearshore of the 305 deforested mangroves were sampled during the ebb to low slack tide, leading to some of the highest 306 DOC concentrations, ranging between 1,000 and 19,400 mol C L-1. Due to tidally-driven 307 variability, no significant differences could be observed in median surface water DOC concentrations 308 between intact and deforested nearshore areas (P = 0.19). 309 310 3.1 Soil physico-chemical properties downcore 311 Changes in soil physico-chemical properties over the soil profile were observed in low-DBD 312 intact and deforested mangrove soils only. In soils from low-DBD intact mangroves, DBD increased 313 linearly and water content decreased with soil cumulative mass (or depth in g cm-2, hereafter). 314 Deforested soils, however, displayed a constant and significantly higher DBD, with decreasing water 315 content over the upper 14 g cm-2 (or ∼ 40 cm) (P< 0.01; Table S2) (Fig. 4a and b). 316 317 Carbon and N content (%DW) in low-DBD intact mangrove soils decreased steadily 318 downcore, opposite to the pattern observed in deforested soils, which increased with depth in the 319 upper 14 g cm-2 (Fig. 4c and d). The mean soil C content of low-DBD intact soils was 8.4 ± 0.2% C 320 in the upper 14 g cm-2, 2-fold higher than that in deforested soils (4.7 ± 0.2% C). Below this horizon, 321 however, no significant differences were observed in water, C or N contents between deforested and 13 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 322 low-DBD intact soils (Fig. 4) (Table S2). In contrast, soil properties of high-DBD intact soils 323 remained relatively constant and differed significantly from those in low-DBD intact and deforested 324 soils throughout the soil profile (Table 2). 325 Molar C:N ratios were high in all sampled soils with averages ranging from 29 to 33, and 326 increased downcore in low-DBD intact and deforested soils. Correlation between C:N ratios and soil 327 cumulative mass was positive and significant in all treatments (P < 0.001) except for the high-DBD 328 intact soils (P > 0.05). The relationship was strongest in the upper 14 g cm2 of deforested soils (rs = 329 0.54) and the slope higher than in the low-DBD intact soils (rs = 0.41) (Fig. S2). Although, 330 deforested soils were characterized by higher average C:N molar ratios over the total soil profile no 331 significant differences were observed in bottom layers between low-DBD intact and deforested soils 332 (Table S2). 333 Stable isotopes of C and N were analysed for one low-DBD intact and one deforested soil. 334 Both mangrove soils showed 13C values close to -28‰ although C in the deforested soil had lighter 335 average 13C values (-28.16 ± 0.05‰) than C in the intact soil (-27.5 ± 0.1‰) (P < 0.001). In 336 contrast, the 15N signal averaged 0.92 ± 0.10‰ and was not significantly different between the two 337 soil types (P = 0.70). Soil C in low-DBD intact soil showed 13C values that became slightly heavier 338 (+1.2‰) over the upper layer and relatively constant in the bottom layer (Fig. 4c). The deforested 339 soil showed an initial change from heavy to lighter 13C values at ~ 3 g cm-2 before becoming 340 heavier downcore (Fig. 4d). 15N showed a similar pattern as 13C in the intact soil but showed 341 scattered values down core in the deforested soil. 342 343 The grain size distribution of all intact and deforested mangrove soils was relatively 344 homogeneous over the soil profile. Silt accounted for 52 ± 5% and 58 ± 6% of dry weight, 345 respectively. However, the average clay content (< 4 m) in deforested soils (28 ± 1%) was about 346 twice that of high-DBD (11 ± 1%) and low- DBD intact soils (15 ± 1%) (Fig. S3). No significant 14 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 347 differences in clay content were observed with depth in any of the sampled soils (P > 0.01; Table 348 S2). 349 350 3.2 210Pb 351 In all intact mangrove soils, the excess 210Pb (210Pbxs) specific activity decreased from the 352 surface to below detection at depths between 7 and 17 g cm-2 (Fig. 5a). The CF:CS model was used 353 to estimate average mass accumulation rates over this depth horizon (Krishnaswamy and others 354 1971), which ranged between 0.070 ± 0.014 g cm-2 y-1 and 0.223 ± 0.012 g cm-2 y-1 (or 0.85 ± 0.11 355 and 8.4 ± 0.4 mm y-1; Table 3) at intact sites. In deforested mangrove soils, 210Pbxs horizons were 356 reached at 10 to 26 g cm-2 (18 ± 3 g cm-2) but intense soil mixing, indicated by uniform specific 357 activities of 210Pbxs throughout deforested soil profiles, precluded the determination of a valid age 358 model and sediment accumulation rates at these sites. Indeed, the cumulative mass of the soil mixed 359 layer was on average 10 times greater in deforested than in intact soils (14 g cm-2 versus 1.4 g cm-2) 360 (Table 3). The 210Pbxs inventories in intact mangrove soils varied widely (600 - 4,800 Bq m-2), 361 although the average (2,210 ± 800 Bq m-2) was not significantly different to that observed in 362 deforested mangrove soils (mean 1,930 ± 220 Bq m-2) (P = 0.74), where the 210Pbxs inventory was 363 redistributed over the soil profile via mixing. It should be noted that high-DBD intact soils contained 364 the lowest 210Pbxs inventories (600 ± 100 Bq m-2). This could potentially mask significant differences 365 between intact and deforested 210Pbxs inventories, and therefore, erosion patterns using the targeted 366 inventory comparison. However, the mean 210Pbxs inventory of low-DBD intact soils was also not 367 significantly different than that of deforested soils (P = 0.08), indicating the lack of erosion at 368 deforested sites relative to reference intact sites. 369 370 3.3 C and N accumulation rates and stocks 15 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 371 Mean soil C and N burial rates during the last century within the closed canopy forest ranged 372 from 18 to 176 g C m-2 y-1 and from 0.7 to 7.2 g N m-2 y-1, respectively, not including belowground 373 biomass accumulation. In low-DBD intact soils, C and N burial rates averaged 110 ± 40 g C m-2 y-1 374 and 4 ± 2 g N m-2 y-1, respectively and were similar to burial rates estimated globally (Breithaupt and 375 others 2012) but 5 times higher than those found in high-DBD intact soils (Table 4). In deforested 376 plots, soil C and N burial rates could not be estimated due to soil mixing over the entire 210Pbxs 377 record. 378 379 Evidence for soil C and N losses was clear between deforested and low-DBD intact soils in 380 the upper 14 g cm-2, which encompasses soils accumulated during the last century, as indicated by the 381 average depth of 210Pbxs horizon in all sampled soils (Table 3). Over this depth and equivalent period 382 of accumulation, stocks of C and N in deforested soils were half those of low-DBD intact soils 383 (Table 4). No significant differences, however, were observed if high-DBD intact soils were included 384 in the comparison or if stock comparisons were made over the upper 45 g cm-2 or 1 m of soil (P > 385 0.05 in all cases). 386 387 4. Discussion 388 Changes in soil C stocks following deforestation can be an important component of the 389 ecosystem C budget and therefore climate mitigation. This is particularly true for ecosystems such as 390 mangrove forests. In Madagascar, clearing and harvesting of the mangrove forest between 1990 and 391 2010 has resulted in an estimated net loss of ~ 21% of mangrove cover (Jones and others 2016a), but 392 the effects of these activities on soil C storage are poorly known. Carbon losses from soils following 393 mangrove clearance are more difficult to assess than aboveground stocks. The slow rate at which soil 394 C stocks change and their inherently large spatial variability make quantification of recent changes 395 unclear, especially when soils have remained on site but have been mixed and/or compacted and 16 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 396 reference soil depths are no longer valid. As evidenced from the low- and high-DBD intact mangrove 397 soils in this study and reported elsewhere (Chmura and others 2003; Ferreira and others 2010; 398 Kauffman and others 2014; Otero and others 2017), intact mangrove soils are heterogeneous. Soil 399 heterogeneity can be increased by mangroves themselves through biotic processes such as 400 colonization, root activity and distribution, and decomposition (Boto and Wellington 1984). It can 401 also occur from environmental processes such as coastal evolution and changes in creek 402 configuration (Macnae 1969; Semeniuk 1996; Ferreira and others 2010), differences in tidal water 403 flooding or sediment supply (Chmura and others 2003). We observed a wide range and variability in 404 soil physico-chemical properties and C and N stocks and accumulation rates in soils from intact 405 mangroves in Tsimipaika Bay that were independent of aboveground biomass, mangrove species or 406 distance from shore, but showed a 5-fold difference in soil accretion rates between soils in high- and 407 low-DBD intact mangrove plots. In contrast, soil properties including 210Pbxs inventories and C and 408 N profiles appeared spatially homogeneous among cleared soils, suggesting that deforestation and 409 the continuous exposure of soils caused a general loss of natural variability. This pattern was also 410 reported by Stoke and Harris (2015) in New Zealand and, although often overlooked, may be an 411 additional sign of ecosystem function and service loss (Stover and Henry 2018). 412 The effects of deforestation and the continuous exposure of soils over 10 years were clear in 413 the upper 14 g cm-2 (or apparent ∼40 cm). Over this depth, all deforested soils showed evidence of 414 disturbance as indicated by lower water content, higher DBD, 10-times higher soil mixing and a 415 higher depletion of C and N contents at the surface. Pre-existing soil properties and C and N stocks 416 were unknown at deforested plots. However, the strong convergence of soil properties between 417 deforested and low-DBD intact soils below 14 g cm-2 (Fig. 4; Table S2) allowed us to use the latter 418 as a reference for pre-deforestation conditions. This approach has been used previously (Grellier and 419 others 2017), and gives first-order estimates of biomass and soil C and N loss. Both low-DBD intact 420 and deforested soils showed isotopic 13C values of sedimentary organic C similar to those of 17 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 421 mangrove vegetation (-29.4 to -27‰; Bouillon and others 2008b) suggesting that the source of 422 organic matter was predominantly of mangrove origin at both sites. The enrichment of 13C and 15N 423 with depth in low-DBD intact soils was consistent with organic matter decomposition with age 424 (Fourqurean and Schrlau 2003) and with the Suess effect, i.e., the temporal decrease in atmospheric 425 CO2 13C signature from the burning of fossil fuels (Keeling 1979). This was not observed in the 426 deforested soil because of sediment mixing in the upper layers. The 10-fold larger soil mixing at 427 deforested sites could have been caused by trampling by harvesters and dragging of logs during 428 clearcutting. As shown in the felling of terrestrial forests, most soil damage occurs during wood 429 transportation from the stump area to the landings (Jamshidi and others 2008; Cambi and others 430 2015). In addition, 10 years of subsequent exposure of reworked soils could have led to increased 431 erosion of cleared soils with tides and run off as observed in Grellier and others (2017) two years 432 after mangrove clearing. However, unlike the study of Grellier et al. (2018) in Vietnam, our data do 433 not show a winnowing of fine particles at deforested sites. Indeed, our deforested soils contained 434 twice as much clay per soil volume as intact soils throughout the soil profile suggesting that their 435 location at the bottom of the U-shaped bay might favor the retention of sediments regardless of 436 mangrove loss. This is confirmed by similar average 210Pbxs inventories measured within deforested 437 and low-DBD intact soils. Despite the large variability at intact sites, results suggest that net soil 438 erosion was not enhanced at deforested sites, hence could not explain the depletion of C and N 439 observed in the upper 14 g cm-2. 440 In our study, the observed C depletion measured in the upper 14 g cm-2 of deforested soils 441 could have occurred largely because of the lack of new litter supply and the enhanced C 442 remineralization promoted by soil physical mixing and the exposure of deforested surface soils to 443 direct solar radiation after canopy loss (Bosire and others 2003; Lovelock and others 2017a). Soil 444 mixing can impact C processing in multiple ways: by aeration, by mixing labile organic matter to 445 deeper layers as well as by breaking the soil structure and exposing organic material previously 18 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 446 protected by burial (Burdige 2007; Middelburg 2018). All these processes promote C mineralization 447 and CO2 emissions to the atmosphere (Lovelock and others 2017a). Assuming that low-DBD intact 448 soils best represent deforested soils before mangrove clearance, we estimated that soil C loss caused 449 by deforestation accounted for 50 ± 14 Mg C ha−1, as indicated by the comparison of C stocks over 450 the upper 14 g cm-2 (Table 4). This C loss would have occurred at a mean rate of 5.0 ± 1.4 Mg C ha-1 451 yr-1 during the 10-yr following clearcutting, which is 4.5 times higher than the annual C 452 sequestration rate estimated in low-DBD intact soils. 453 Roughly 20% of the upper 1-m C stock, and 45% of that accumulated in the last century, 454 would have been lost since deforestation relative to low-DBD intact soils. The magnitude of C loss 455 in the upper 14 g cm-2 was, however, as important as the variability of the C stocks in the upper 45 g 456 cm-2 (apparent ∼1 m) of low-DBD and deforested soils. This may explain why differences in soil C 457 stocks at a depth of 1 meter could not be detected in this study and others (Lang’at and others 2014), 458 unless specific soil mass-depth increments that encompass similar accumulation periods were 459 analyzed. Proportional changes in N stocks were comparable. Results indicated a preferential loss of 460 N with age (or depth) under natural conditions that was enhanced with disturbance as deforested 461 soils showed a greater increase in average C:N ratios than intact soils throughout the upper 14 cm-2 462 (Fig. S2). Soil C loss occurred in addition to the loss from standing biomass, which was estimated at 463 130 ± 14 Mg C ha-1 (Fig. 6). Soil C and N losses (versus tree biomass) could have been lower if 464 deforestation had occurred for mangroves in high-DBD intact plots given their smaller organic C and 465 N soil content (Lovelock and others 2017a). 466 467 Pathways for soil C loss likely include atmospheric emissions as CO2 and lateral export as 468 dissolved C to coastal waters. Most research has focused on constraining C losses from soils as 469 fluxes of CO2 following disturbance (e.g., references in Table 5). For comparison, the mean C loss 470 rate estimated in deforested soils here is equivalent to emissions of 18 ± 5 Mg CO2 ha-1 y-1, which 19 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 471 compare well with CO2 efflux measurements after mangrove clearing reported by others (e.g., 472 Lang’at and others 2014; Bulmer and others 2015; Grellier and others 2017) and is similar to CO2 473 emissions inferred from peat collapse due to hurricane damage (Cahoon and others 2003) and from 474 stock change methods after conversion to cattle pastures (Kauffman and others 2016) (Table 5). 475 However, our estimated emissions were about 4 times lower than those reported when soils were 476 excavated and converted to shrimp ponds (Sidik and Lovelock 2013; Kauffman and others 2014, 477 2018). 478 Lateral fluxes of dissolved C from cleared mangroves have been rarely considered (Sippo and 479 others 2019), despite this being the major fate of C from healthy mangrove forests (Bouillon and 480 others 2008; Maher and others 2018). Although our data were limited, DOC concentrations of waters 481 nearshore of the deforested and intact forests were very high, even during flood tide, far exceeding 482 those typical of seawater and coastal waters (∼100 and 200 mol C L-1, respectively; Barrón and 483 Duarte 2015), which confirms the importance of mangrove forests, whether intact or degraded, as a 484 source of DOC to the coastal ocean. The lack of samples taken along river sources and coastal areas 485 at the same tidal stage precluded assessing the effect of mangrove deforestation on the magnitude of 486 DOC export. However, because mangrove deforestation limits the supply of new C inputs to the 487 forest floor, the presence of high DOC concentrations (median: 2,000 mol C L-1) in waters 488 nearshore of the cleared area even 10 years after deforestation may suggest that existing soil C 489 reserves are being depleted. This is consistent with the findings by Maher and others (2017), who 490 showed that even aged sequestered C in mangroves is susceptible to remineralization and export to 491 the coastal ocean. This process reintroduces aged C into the modern C cycle and thus could lead to 492 increased CO2 emissions in coastal waters if bioavailable and oxidized (Drake and others 2019). 493 494 Mangrove deforestation promotes changes in soil physico-chemical properties and functions 495 (soil C storage, nutrient processing and vertical accretion) (Grellier and others 2017; Otero and 20 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 496 others 2017) that enhance the susceptibility of C stocks to remineralization, resulting in potential 497 changes in C fluxes to the coast and large amounts of CO2 to the atmosphere. Using “tier 2” 498 approaches, we estimate that mangrove deforestation for timber and charcoal in Tsimipaika Bay has 499 resulted in measurable reductions in total ecosystem C stocks that represents a combined potential 500 loss of 180 ± 20 Mg C ha-1 from standing biomass and soil organic C stocks in the 10 years since 501 clearing. While C losses from standing biomass are unequivocal and could contribute significantly to 502 CO2 emissions if harvested timber is used as fuel-wood, emissions from soils are more difficult to 503 assess because of the large soil heterogeneity of intact mangrove forests and the partial export of C to 504 adjacent aquatic coastal systems. However, our data show that even in the absence of excavation of 505 soils (e.g. for aquaculture ponds, Kauffman and others 2014) or soil erosion, C losses would have 506 occurred at a rate that is 4.5 times that which C accumulates in soils of intact closed-canopy forests. 507 In Tsimipaika-Ambaro Bay, closed-canopy mangrove forests cover 14,000 ha in contrast to 1,000 ha 508 of deforested soils, which thereby reduce the annual C sequestration capacity of the dense mangrove 509 ecosystem by 32%. Nation-wide, Madagascar has lost 20,300 ha of mangroves between 2000 and 510 2010 (Jones and others 2016a). If our results are broadly representative, C loss from cleared 511 mangrove soils could be significant and account for approximately 20% of total national C emissions 512 from fossil fuel combustion over that 10 year period (Boden and others 2017). Although rates of 513 mangrove deforestation have slowed in the last decade (Hamilton and Casey 2016), deforestation of 514 mangroves is still a nationally important source of emissions for many nations, particularly those 515 with high rates of deforestation and moderate emissions from other sectors (Taillardat and others 516 2018). Our results show the importance of avoiding CO2 emissions associated with mangrove 517 deforestation, particularly, given that establishing high rates of carbon uptake through restoration of 518 mangroves can take decades (Osland and others 2012). Thus, conservation is an effective mechanism 519 for reducing CO2 emissions. Conservation projects seeking to account for C emissions and removals 21 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 520 should take into consideration avoided emissions from soils as well as from loss of biomass, even in 521 cases where excavation of soils has not taken place. 522 523 Data availability 524 Data supporting the findings of this study (DBD, water content, grains size distribution, C and N 525 contents, 13C and 15N, 210Pb and DOC) are available at https://ddd.uab.cat/record/216456 with the 526 identifier doi:10.5565/ddd.uab.cat/216456 527 528 Acknowledgements 529 We thank the Blue Ventures team in Madagascar and Samantha Ridgway and Gloria Salgado from 530 Edith Cowan University for their help in the field and lab work. Funding was provided to PM and 531 JGO by the Generalitat de Catalunya (Grant 2017 SGR-1588) and to PM through an Australian 532 Research Council LIEF Project (LE170100219). Funding was provided to MH and CBN through a 533 South African NRF Research Development grant (Grant No: 105724). Blue Ventures was funded by 534 the GEF Blue Forests Project. This work is contributing to the ICTA ‘Unit of Excellence’ (MinECo, 535 MDM2015-0552). AA-O was supported by “Obra Social la Caixa” (LCF/BQ/ES14/10320004) and 536 by the NOAA C&GC Postdoctoral Fellowship Program administered by UCAR-CPAESS under 537 award #NA18NWS4620043B. 538 539 References 540 Adame MF, Zakaria RM, Fry B, Chong VC, Then YHA, Brown CJ, Lee SY. 2018. Loss and recovery of 541 carbon and nitrogen after mangrove clearing. Ocean Coast Manag 161:117–26. 542 Alongi DM. 2002. Present state and future of the world’s mangrove forests. Environ Conserv 29:331–49. 22 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 543 Appleby PG. 2001. Chronostratigraphic Techniques in Recent Sediments. In: Tracking Environmental Change 544 Using Lake Sediments. Vol. 1. Springer Netherlands. pp 171–203. 545 Arata L, Meusburger K, Frenkel E, A’Campo-Neuen A, Iurian AR, Ketterer ME, Mabit L, Alewell C. 2016. 546 Modelling Deposition and Erosion rates with RadioNuclides (MODERN) - Part 1: A new conversion 547 model to derive soil redistribution rates from inventories of fallout radionuclides. J Environ Radioact 548 162–163:45–55. 549 Arias-Ortiz A, Masqué P, Garcia-Orellana J, Serrano O, Mazarrasa I, Marbà N, Lovelock CE, Lavery P, 550 Duarte CM. 2018. Reviews and syntheses: 210Pb-derived sediment and carbon accumulation rates in 551 vegetated coastal ecosystems: setting the record straight. Biogeosciences 15:6791–818. 552 Atwood TB, Connolly RM, Almahasheer H, Carnell PE, Duarte CM, Ewers Lewis CJ, Irigoien X, Kelleway 553 JJ, Lavery PS, Macreadie PI, Serrano O, Sanders CJ, Santos I, Steven ADL, Lovelock CE. 2017. Global 554 patterns in mangrove soil carbon stocks and losses. Nat Clim Chang 7:523–8. 555 Balesdent J, Chenu C, Balabane M. 2000. Relationship of soil organic matter dynamics to physical protection 556 and tillage. Soil Tillage Res 53:215–30. 557 Barbier EB, Hacker SD, Kennedy C, Koch EW, Stier AC, Silliman BR. 2011. The value of estuarine and 558 coastal ecosystem services. Ecol Monogr 81:169–93. 559 Barrón C, Duarte CM. 2015. Global Biogeochemical Cycles from the coastal ocean. Global Biogeochem 560 Cycles 29:1725–38. 561 Benner R, Strom M. 1993. A critical evaluation of the analytical blank associated with. MarChem 41:153–60. 562 Benson L, Glass L, Jones TG, Ravaoarinorotsihoarana L, Rakotomahazo C. 2017. Mangrove carbon stocks 563 and ecosystem cover dynamics in southwest Madagascar and the implications for local management. 564 Forests 8:1–21. 565 Bianchi TS. 2011. The role of terrestrially derived organic carbon in the coastal ocean: A changing paradigm 566 and the priming effect. Proc Natl Acad Sci 108:19473–81. 23 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 567 Boden TA, Marland G, Andres RJ. 2017. Global, Regional, and National Fossil-Fuel CO2 Emissions. 568 https://cdiac.ess-dive.lbl.gov/trends/emis/overview_2014.html. Last accessed 13/10/2019 569 Bosire JO, Dahdouh-Guebas F, Kairo JG, Koedam N. 2003. Colonization of non-planted mangrove species 570 into restored mangrove stands in Gazi Bay, Kenya. Aquat Bot 76:267–79. 571 Boto K, Wellington J. 1984. Soil characteristics and nutrient status in a northern Australian mangrove forest. 572 Estuaries. 573 Bouillon S, Borges A V., Castañeda-Moya E, Diele K, Dittmar T, Duke NC, Kristensen E, Lee SY, Marchand 574 C, Middelburg JJ, Rivera-Monroy VH, Smith TJ, Twilley RR. 2008. Mangrove production and carbon 575 sinks: A revision of global budget estimates. Global Biogeochem Cycles 22:1–12. 576 Breithaupt JL, Smoak JM, Smith TJ, Sanders CJ, Hoare A. 2012. Organic carbon burial rates in mangrove 577 sediments: Strengthening the global budget. Global Biogeochem Cycles 26. 578 Bulmer RH, Lundquist CJ, Schwendenmann L. 2015. Sediment properties and CO2 efflux from intact and 579 cleared temperate mangrove forests. Biogeosciences 12:6169–80. 580 Burdige DJ. 2007. Preservation of organic matter in marine sediments: Controls, mechanisms, and an 581 imbalance in sediment organic carbon budgets? Chem Rev 107:467–85. 582 Cahoon DR, Hensel P, Rybczyk J, McKee KL, Proffitt CE, Perez BC. 2003. Mass tree mortality leads to 583 mangrove peat collapse at Bay Islands, Honduras after Hurricane Mitch. J Ecol 91:1093–105. 584 Cambi M, Certini G, Neri F, Marchi E. 2015. The impact of heavy traffic on forest soils: A review. For Ecol 585 Manage 338:124–38. 586 Chave J, Andalo C, Brown S, Cairns MA, Chambers JQ, Eamus D, Fölster H, Fromard F, Higuchi N, Kira T, 587 Lescure J-P, Nelson BW, Ogawa H, Puig H, Riéra B, Yamakura T. 2005. Tree allometry and improved 588 estimation of carbon stocks and balance in tropical forests. Oecologia 145:87–99. 589 Chave J, Coomes D, Jansen S, Lewis SL, Swenson NG, Zanne AE. 2009. Towards a worldwide wood 590 economics spectrum. Ecol Lett 12:351–66. 24 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 591 Chmura GL, Anisfeld SC, Cahoon DR, Lynch JC. 2003. Global carbon sequestration in tidal, saline wetland 592 soils. Global Biogeochem Cycles 17:1111. 593 Clough BF, Scott K. 1989. Allometric relationships for estimating above-ground biomass in six mangrove 594 species. For Ecol Manage 27:117–27. 595 Cole TG, Ewel KC, Devoe NN. 1999. Structure of mangrove trees and forests in Micronesia. For Ecol 596 Manage 117:95–109. 597 Delgado M, de Jonge VN, Peletier H. 1991. Experiments on resuspension of natural microphytobenthos 598 populations. Mar Biol 108:321–8. 599 Diochon A, Kellman L, Beltrami H. 2009. Looking deeper: An investigation of soil carbon losses following 600 harvesting from a managed northeastern red spruce (Picea rubens Sarg.) forest chronosequence. For Ecol 601 Manage 257:413–20. 602 Donato DC, Kauffman JB, Murdiyarso D, Kurnianto S, Stidham M. 2011. Mangroves among the most 603 carbon-rich forests in the tropics. Nat Geosci 4:1–5. 604 Drake TW, Van Oost K, Barthel M, Bauters M, Hoyt AM, Podgorski DC, Six J, Boeckx P, Trumbore SE, 605 Cizungu Ntaboba L, Spencer RGM. 2019. Mobilization of aged and biolabile soil carbon by tropical 606 deforestation. Nat Geosci 12:541–6. 607 Duke N, Wolanski E. 2001. Muddy Coastal Waters and Depleted Mangrove Coastlines — Depleted Seagrass 608 and Coral Reefs. In: Wolanski E, editor. Oceanographic Processes of Coral Reefs. Physical and Biology 609 Links in the Great Barrier Reef. Washington, DC: CRC Press. pp 77–91. 610 Ellison AM, Farnsworth EJ. 1996. Anthropogenic Disturbance of Caribbean Mangrove Ecosystems: Past 611 Impacts, Present Trends, and Future Predictions. Biotropica 28:549. 612 FAO [Food and Agriculture Organization of the United Nations]. 2007. The world’s mangroves 1980-2005. 613 Rome, Italy 614 Ferreira TO, Otero XL, de Souza Junior VS, Vidal-Torrado P, Macías F, Firme LP. 2010. Spatial patterns of 25 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 615 soil attributes and components in a mangrove system in Southeast Brazil (São Paulo). J Soils Sediments 616 10:995–1006. 617 Fourqurean JW, Schrlau JE. 2003. Changes in nutrient content and stable isotope ratios of C and N during 618 decomposition of seagrasses and mangrove leaves along a nutrient availability gradient in Florida Bay, 619 USA. Chem Ecol 19:373–90. 620 Gifford RM, Roderick ML. 2003. Soil carbon stocks and bulk density: Spatial or cumulative mass coordinates 621 as a basis of expression? Glob Chang Biol 9:1507–14. 622 Gillis LG, Belshe EF, Narayan GR. 2017. Deforested Mangroves Affect the Potential for Carbon Linkages 623 between Connected Ecosystems. Estuaries and Coasts 40:1207–13. 624 Glew JR, Smol JP, Last WM. 2001. Sediment Core Collection and Extrusion. In: Last WM, Smol JP, editors. 625 Tracking Environmental Change Using Lake Sediments: Basin Analysis, Coring, and Chronological 626 Techniques. Dordrecht: Springer Netherlands, Dordrecht. pp 73–105. 627 Granek E, Ruttenberg BI. 2008. Changes in biotic and abiotic processes following mangrove clearing. Estuar 628 Coast Shelf Sci 80:555–62. 629 Grellier S, Janeau JL, Dang Hoai N, Nguyen Thi Kim C, Le Thi Phuong Q, Pham Thi Thu T, Tran-Thi NT, 630 Marchand C. 2017. Changes in soil characteristics and C dynamics after mangrove clearing (Vietnam). 631 Sci Total Environ 593–594:654–63. 632 Haines WB. 1923. The volume-changes associated with variations of water content in soil. J Agric Sci 633 13:296–310. 634 Hamilton SE, Casey D. 2016. Creation of a high spatio-temporal resolution global database of continuous 635 mangrove forest cover for the 21st century (CGMFC-21). Glob Ecol Biogeogr 25:729–38. 636 Hargis TG, Twilley RR. 1994. Improved coring device for measuring soil bulk density in a Louisiana deltaic 637 marsh. J Sediment Res 64:681–3. 638 Herr D, von Unger M, Laffoley D, McGivern A. 2017. Pathways for implementation of blue carbon 26 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 639 initiatives. Aquat Conserv Mar Freshw Ecosyst 27:116–29. 640 IPCC. 2014. 2013 Supplement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories: 641 Wetlands Task Force on National Greenhouse Gas Inventories: Wetlands. (Hiraishi T, Krug T, Kiyoto T, 642 Srivastava N, Jamsranjav B, Fukuda M, Troxler T, editors.). Switzerland: IPCC 643 Jamshidi R, Jaeger D, Raafatnia N, Tabari M. 2008. Influence of Two Ground-Based Skidding Systems on 644 Soil Compaction Under Different Slope and Gradient Conditions. Int J For Eng 19:9–16. 645 Jones TG, Glass L, Gandhi S, Ravaoarinorotsihoarana L, Carro A, Benson L, Ratsimba HR, Giri C, 646 Randriamanatena D, Cripps G. 2016a. Madagascar’s mangroves: Quantifying nation-wide and 647 ecosystem specific dynamics, and detailed contemporary mapping of distinct ecosystems. Remote Sens 648 8. 649 Jones TG, Ratsimba HR, Carro A, Ravaoarinorotsihoarana L, Glass L, Teoh M, Benson L, Cripps G, Giri C, 650 Zafindrasilivonona B, Raherindray R, Andriamahenina Z, Andriamahefazafy M. 2016b. The Mangroves 651 of Ambanja and Ambaro Bays, Northwest Madagascar: Historical Dynamics, Current Status and 652 Deforestation Mitigation Strategy Trevor. In: Diop S et al., editor. Estuaries: A Lifeline of Ecosystem 653 Services in the Western Indian Ocean. Springer International Publishing Switzerland. pp 67–85. 654 Jones TG, Ratsimba HR, Ravaoarinorotsihoarana L, Cripps G, Bey A. 2014. Ecological Variability and 655 Carbon Stock Estimates of Mangrove Ecosystems in Northwestern Madagascar. Forests 5:177–205. 656 Kauffman JB, Bernardino AF, Ferreira TO, Bolton NW, Eduardo L, Gabriel DOG, Nobrega N. 2018. Shrimp 657 ponds lead to massive loss of soil carbon and greenhouse gas emissions in northeastern Brazilian 658 mangroves. Ecol Evol 8:5530–40. 659 Kauffman JB, Donato DC. 2012. Protocols for the measurement, monitoring and reporting of structure, 660 biomass and carbon stocks in mangrove forests. Bogor, Indonesia: CIFOR 661 Kauffman JB, Heider C, Norfolk J, Payton F. 2014. Carbon stocks of intact mangroves and carbon emissions 662 arising from their conversion in the Dominican Republic. Ecol Appl 24:518–27. 663 Kauffman JB, Hernandez Trejo H, del Carmen Jesus Garcia M, Heider C, Contreras WM. 2016. Carbon 27 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 664 stocks of mangroves and losses arising from their conversion to cattle pastures in the Pantanos de Centla, 665 Mexico. Wetl Ecol Manag 24:203–16. 666 Kauffman JB, Thorpe AS, Brookshire ENJ. 2004. Livestock exclusion and belowground ecosystem responses 667 in riparian meadows of Eastern Oregon. Ecol Appl 14:1671–9. 668 Keeling CD. 1979. The Suess effect: 13Carbon-14Carbon interrelations. Environ Int 2:229–300. 669 Kennedy P, Kennedy H, Papadimitriou S. 2005. The effect of acidification on the determination of organic 670 carbon, total nitrogen and their stable isotopic composition in algae and marine sediment. Rapid 671 Commun Mass Spectrom 19:1063–8. 672 Komiyama A, Poungparn S, Kato S. 2005. Common allometric equations for estimating the tree weight of 673 mangroves. J Trop Ecol 21:471–7. 674 Krauss KW, Cahoon DR, Allen JA, Ewel KC, Lynch JC, Cormier N. 2010. Surface Elevation Change and 675 Susceptibility of Different Mangrove Zones to Sea-Level Rise on Pacific High Islands of Micronesia. 676 Ecosystems 13:129–43. 677 Krishnaswamy S, Lal D, Martin JM, Meybeck M. 1971. Geochronology of lake sediments. Earth Planet Sci 678 Lett 11:407–14. 679 Kristensen E, Alongi DM. 2006. Control by fiddler crabs ( Uca vocans ) and plant roots ( Avicennia marina ) 680 on carbon, iron, and sulfur biogeochemistry in mangrove sediment. Limnol Oceanogr 51:1557–71. 681 Labrière N, Locatelli B, Laumonier Y, Freycon V, Bernoux M. 2015. Soil erosion in the humid tropics: A 682 systematic quantitative review. Agric Ecosyst Environ 203:127–39. 683 Lang’at JKS, Kairo JG, Mencuccini M, Bouillon S, Skov MW, Waldron S, Huxham M. 2014. Rapid losses of 684 surface elevation following tree girdling and cutting in tropical mangroves. PLoS One 9:1–8. 685 Lovelock CE, Atwood T, Baldock J, Duarte CM, Hickey S, Lavery PS. 2017a. Assessing the risk of carbon 686 dioxide emissions from blue carbon ecosystems. Front Ecol Environ 15:257–65. 687 Lovelock CE, Fourqurean JW, Morris JT. 2017b. Modeled CO2 Emissions from Coastal Wetland Transitions 28 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 688 to Other Land Uses: Tidal Marshes, Mangrove Forests, and Seagrass Beds. Front Mar Sci 4:1–11. 689 Lovelock CE, Ruess R, Feller I. 2011. CO2 efflux from cleared mangrove peat. PLoS One 6:e21279. 690 Lundquist CJ, Morrisey DJ, Gladstone-Gallagher R V., Swales A. 2014. Managing Mangrove Habitat 691 Expansion in New Zealand. In: Mangrove Ecosystems of Asia. New York, NY: Springer New York. pp 692 415–38. 693 Macnae W. 1969. A General Account of the Fauna and Flora of Mangrove Swamps and Forests in the Indo- 694 West-Pacific Region. Adv Mar Biol 6:73–270. 695 Maher DT, Call M, Santos IR, Sanders CJ. 2018. Beyond burial: Lateral exchange is a significant atmospheric 696 carbon sink in mangrove forests. Biol Lett 14. 697 Maher DT, Santos IR, Schulz KG, Call M, Jacobsen GE, Sanders CJ. 2017. Blue carbon oxidation revealed by 698 radiogenic and stable isotopes in a mangrove system. Geophys Res Lett 44:4889–96. 699 McKee KL. 2011. Biophysical controls on accretion and elevation change in Caribbean mangrove 700 ecosystems. Estuar Coast Shelf Sci 91:475–83. 701 McLeod E, Chmura GL, Bouillon S, Salm R, Björk M, Duarte CM, Lovelock CE, Schlesinger WH, Silliman 702 BR. 2011. A blueprint for blue carbon: Toward an improved understanding of the role of vegetated 703 coastal habitats in sequestering CO2. Front Ecol Environ 9:552–60. 704 Middelburg JJ. 2018. Reviews and syntheses: To the bottom of carbon processing at the seafloor. 705 Biogeosciences 15:413–27. 706 Morton RA, White WA. 1997. Characteristics of and corrections for core shortening in unconsolidated 707 sediments. J Coast Res 13:761–9. 708 Ong JE. 1993. Mangroves - a carbon source and sink. Chemosphere 27:1097–107. 709 Osland MJ, Spivak AC, Nestlerode JA, Lessmann JM, Almario AE, Heitmuller PT, Russell MJ, Krauss KW, 710 Alvarez F, Dantin DD, Harvey JE, From AS, Cormier N, Stagg CL. 2012. Ecosystem Development 711 After Mangrove Wetland Creation: Plant-Soil Change Across a 20-Year Chronosequence. Ecosystems 29 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 712 15:848–66. 713 Otero XL, Méndez A, Nóbrega GN, Ferreira TO, Meléndez W, Macías F. 2017. High heterogeneity in soil 714 composition and quality in different mangrove forests of Venezuela. Environ Monit Assess 189. 715 Rasolofo M, Ramilijaona O. 2009. Variability in the abundance and recruitment of Fenneropenaeus indicus 716 and Metapenaeus monoceros postlarvae and juveniles in Ambaro Bay mangroves of Madagascar. Nat 717 Faune 24:103–9. 718 Sanchez-Cabeza JA, Masqué P, Ani-Ragolta I. 1998. 210Pb and210Po analysis in sediments and soils by 719 microwave acid digestion. J Radioanal Nucl Chem 227:19–22. 720 Semeniuk V. 1996. Coastal forms and Quaternary processes along the arid Pilbara coast of northwestern 721 Australia. Palaeogeogr Palaeoclimatol Palaeoecol 123:49–84. 722 Sidik F, Lovelock CE. 2013. CO2 Efflux from Shrimp Ponds in Indonesia. PLoS One 8:6–9. 723 Sippo JZ, Maher DT, Schulz KG, Sanders CJ, McMahon A, Tucker J, Santos IR. 2019. Carbon outwelling 724 across the shelf following a massive mangrove dieback in Australia: Insights from radium isotopes. 725 Geochim Cosmochim Acta 253:142–58. 726 Stover HJ, Henry HAL. 2018. Soil homogenization and microedges: perspectives on soil-based drivers of 727 plant diversity and ecosystem processes. Ecosphere 9:e02289. 728 Taillardat P, Friess DA, Lupascu M. 2018. Mangrove blue carbon strategies for climate change mitigation are 729 most effective at the national scale. Biol Lett 14:20180251. 730 Thampanya U, Vermaat JE, Sinsakul S, Panapitukkul N. 2006. Coastal erosion and mangrove progradation of 731 Southern Thailand. Estuar Coast Shelf Sci 68:75–85. 732 Twilley RR, Day JW. 2012. Mangrove Wetlands. In: W. Day J, C. Crump B, Kemp WM, Yáñez‐Arancibia A, 733 editors. Estuarine Ecology. Second. Hoboken, NJ, USA: John Wiley & Sons, Inc. pp 165–202. 734 Wendt JW, Hauser S. 2013. An equivalent soil mass procedure for monitoring soil organic carbon in multiple 735 soil layers. Eur J Soil Sci 64:58–65. 30 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 736 Wentworth C. 1922. A scale of grade and class terms for clastic sediments. J Geol 30:377–92. 737 Yanai RD, Currie WS, Goodale CL. 2003. Soil carbon dynamics after forest harvest: An ecosystem paradigm 738 reconsidered. Ecosystems 6:197–212. 739 Zanne AE, Lopez-Gonzalez G, Coomes DA, Ilic J, Jansen S, Lewis SL, Miller RB, Swenson NG, Wiemann 740 MC, Chave J. 2009. Data from: towards a worldwide wood economics spectrum. Dryad Digit Repos. 741 Zummo LM, Friedland AJ. 2011. Soil carbon release along a gradient of physical disturbance in a harvested 742 northern hardwood forest. For Ecol Manage 261:1016–26. 743 31 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 744 Table Legends 745 Table 1. Site characteristics for sampled mangrove plots of Tsimipaika Bay, Madagascar. 746 Average values for tree height (m), diameter at breast height (dbh) (cm) and trees per hectare (ha-1). 747 Dead tree density in deforested plots equals stump density and regeneration density equals number of 748 seedlings per unit area. Cc stands for Closed-canopy, De for Deforested, and n.d. for “no data”. 749 750 Table 2. Intact and deforested soil characteristics over the upper 45 g·cm-2. U and t values of 751 Mann-Witney and Two-sample t- test results are included for the comparison of soil properties 752 between intact and deforested mangrove soils. ***Significant at 0.001 level, ** at 0.01 and * at 0.05. 753 NS is not significant. 754 755 Table 3. Sedimentation rates and 210Pbxs inventories in intact and deforested soils. The 756 uncertainties represent the SE resulting from the CF:CS model to obtain the mass accumulation and 757 accretion rates (MAR and SAR), and 226Ra specific activities represent the mean and the standard 758 deviation (n = 5 at each core). Accretion rates (SAR) were corrected for core shortening thus should 759 be considered as apparent rates. 760 761 Table 4. Soil carbon and nitrogen accumulation rates and stocks. Stocks in the upper 1 meter are 762 also included for reporting purposes and should be considered apparent stocks due to the soil 763 shortening correction applied. 764 765 Table 5. Published C loss rates from degraded mangrove soils. To facilitate comparison among 766 most other assessments and this study, C losses are expressed as CO2 equivalents (CO2e) obtained by 767 multiplying C loss rate by 3.67. Where rates of C loss were not reported, we estimated a mean annual 768 C loss as the total soil C stock loss divided by the time since disturbance. 32 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 769 Table 1. Live Dead Tree Total Geomorphic Tidal dbh Tree Tree Regeneration Canopy Core ID Species dominance height biomass C position inundation (cm) Density Density density (ha-1) cover (%) (m) (Mg C ha-1) (ha-1) (ha-1) Cc18 S. alba Riverine freq. 8 18 800 0 0 94 160 Cc19 B. gymnorhiza Riverine infreq. 10 13 2400 200 5200 95 254 Cc20 R. mucronata Basin freq. 9 10 4700 200 800 95 238 Cc28 R. mucronata Basin infreq. n.d. 10 2900 400 2800 93 128 Cc29 R. mucronata Riverine freq. 9 9 2900 600 0 98 113 De27 B. gymnorhiza Riverine freq. 5 7 200 100 0 0.9 3.0 De30 B. gymnorhiza Basin infreq. 0 1000 2000 0.2 1.8 De31 R. mucronata Basin freq. 0 100 0 0.2 0.1 De32 C. tagal Riverine infreq. 0 3000 2400 7 1.4 De33 C. tagal Basin infreq. 0 11400 4400 0.2 9.5 770 771 Table 2. Core Water Soil Class Statistic DBD (g cm-3) C (%DW) N (%DW) C:N Clay (%) ID content (%) High- Mean ± SE 30.1 ± 0.6 0.881 ± 0.013 2.40 ± 0.08 0.095 ± 0.003 28.6 ± 0.6 11.2 ± 1.0 Cc19, DBD 20 Median 29.1 0.88 2.3 0.093 28.0 10.3 Intact Cc18, Mean ± SE 56.9 ± 0.7 0.347 ± 0.008 7.7 ± 0.2 0.314 ± 0.012 29.6 ± 0.4 15.3 ± 1.1 Low-DBD 28, Intact Median 58.7 0.36 7.4 0.30 28.9 15 29 Cc18, Mean ± SE 47.4 ± 0.9 0.580 ± 0.015 5.7 ± 0.2 0.231 ± 0.011 29.3 ± 0.3 14.0 ± 0.8 19, Intact (all) 20, 28, Median 48.7 0.49 5.5 0.20 28.8 13.1 29 De27, Mean ± SE 44.5 ± 0.4 0.415 ± 0.006 5.11 ± 0.15 0.173 ± 0.003 33.5 ± 0.6 28.1 ± 1.3 30, Deforeste 31, d Median 44.3 0.41 4.8 0.17 32.6 26.4 32, 33 Treatment Prob>|U| Prob>|t| Intact (all) vs. Deforested 0.007 ** 1.1·10-9 *** 1.00 NS 0.11 NS 2.6·10-8*** 1.1·10-14 *** High-DBD Intact vs. Deforested 0 *** 5.6·10-53*** 4.1·10-26*** 1.7·10-25*** 7.0·10-6*** 9.4·10-10 *** Low-DBD Intact vs. Deforested 0 *** 2.6·10-9*** 0*** 0*** 2.6·10-6*** 3.4·10-10 *** 772 773 774 33 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 775 Table 3. 210 Core Mixing Pbxs 226 210 Type Ra MAR SAR Pbxs inventory ID depth horizon g cm-2 g cm-2 Bq kg-1 g cm-2 yr-1 mm yr-1 Bq m-2 Cc19 1.7 7 13.2 ± 1.5 0.070 ± 0.014 0.85 ± 0.11 500 ± 50 High-DBD Cc20 2.0 9 12.6 ± 1.3 0.099 ± 0.013 1.11 ± 0.15 690 ± 40 Cc18 0.8 11 19 ± 4 0.094 ± 0.007 4.1 ± 0.2 3170 ± 90 Cc28 Low-DBD 1.2 13 13 ± 4 0.10 ± 0.02 2.3 ± 0.4 1920 ± 80 Cc29 1.5 17 13 ± 2 0.223 ± 0.012 8.4 ± 0.4 4750 ± 80 Mean (SE) high-DBD 1.9 ± 0.2 8.0 ± 0.8 0.085 ± 0.015 0.98 ± 0.13 600 ± 100 Mean (SE) low-DBD 1.2 ± 0.2 13 ± 2 0.14 ± 0.04 5 ± 2 3280 ± 820 Mean (SE) Intact 1.4 ± 0.2 11 ± 2 0.12 ± 0.03 3.4 ± 1.4 2210 ± 800 De27 4 14 25 ± 6 1940 ± 90 De30 6 14 18 ± 4 1200 ± 70 De31 Deforested 25 24 13.0 ± 1.4 2580 ± 90 De32 8 10 21 ± 6 1870 ± 170 De33 26 26 15 ± 2 2040 ± 100 Mean (SE) Deforested 14 ± 5 18 ± 3 1930 ± 220 776 34 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 777 Table 4. Core ID Type C accumulation rate N accumulation rate C stock N stock C stock N stock 0-14 g cm-2 0-45 g cm-2 0-14 g cm-2 0-45 g cm-2 1m 1m gC m-2 yr-1 gN m-2 yr-1 Mg C ha-1 Mg N ha-1 Mg C ha-1 Mg N ha-1 Cc19 18 ± 4 0.68 ± 0.14 32 109 1.3 4.1 218 8 High-DBD Cc20 25 ± 3 1.18 ± 0.15 33 91 1.5 3.9 192 8 Cc18 75 ± 6 3.4 ± 0.3 108 279 4.4 11 190 7 Cc28 Low-DBD 64 ± 11 2.2 ± 0.4 87 248 3.1 8.9 250 9 Cc29 176 ± 9 7.3 ± 0.4 135 305 5.8 13 229 9 Mean (SE) High-DBD Intact 21 ± 3 0.9 ± 0.3 32.4 ± 0.4 100 ± 9 1.40 ± 0.12 3.99 ± 0.06 Mean (SE) Low-DBD Intact 110 ± 40 4 ± 2 110 ± 14 280 ± 20 4.4 ± 0.8 11.0 ± 1.1 Mean (SE) Intact 70 ± 30 3.0 ± 1.2 80 ± 20 200 ± 40 2.9 ± 0.8 8±2 220 ± 10 8.3 ± 0.3 De27 64 268 2.1 8 268 8 De30 69 344 2.2 10 279 8 De31 Deforested 53 194 2.1 7 181 6 De32 57 261 3.1 10 166 6 De33 52 170 2.2 7 135 5 Mean (SE) deforested 60 ± 3 250 ± 30 2.3 ± 0.2 8.2 ± 0.7 210 ± 30 7.0 ± 0.5 778 35 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 779 Table 5. CO2 emissions Years since Method for estimating C References Disturbance type disturbance loss Mg CO2e ha-1 yr- 1 This study Clearing 10 C stock change 18 ± 5 Grellier and others C stock change 37 Clearing 2 (2017) Gas flux chambers 4 ± 7 Bulmer and others Clearing 0.1-8 Gas flux chambers 21 ± 6 (2015) Lang’at and others C stock change 35 ± 45 Clearing 2 (2014) Gas flux chambers 25 ± 7 Lovelock and others 1 106 Clearing Gas flux chambers (2011) 20 30 Kauffman and others Conversion to cattle 7 16 ± 6 C stock change (2016b) pastures 30 7 ± 2 Kauffman and others Conversion to 29 C stock change 82 (2014) aquaculture 10-12 107 ± 40 Kauffman and others Conversion to 10-12 C stock change 184 ± 10 (2018) aquaculture 8 13 ± 5 Gas flux chambers in pond 16 Sidik and Lovelock Conversion to floor 25 (2013) aquaculture Gas flux chambers in pond 44 walls Cahoon and others Hurricane damage 2 Change in soil volume 19 (2003) 780 781 36 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 782 Figure Legends 783 Figure 1. Map of Tsimipaika Bay in northwest Madagascar with sampled plot locations in 784 intact and deforested mangrove areas. St. labels are surface water sampling locations. 785 786 Figure 2. Principal component analysis on physico-chemical properties of soils from 787 deforested and intact mangroves. Biplot of variable vectors showing correlation between the 788 variables, the component and individual factor map. Superimposed on the plot are the 789 confidence ellipses for categorical variables: deforested soils (orange circles), low-DBD 790 intact mangrove soils (grey squares) and high-DBD intact mangrove soils (black triangles). 791 792 Figure 3. Concentrations of DOC in surface water at 8 stations along the shores of the intact 793 and deforested mangrove areas. 794 795 Figure 4. Soil properties (bulk density, water, carbon and nitrogen contents) with cumulative 796 mass in intact and deforested mangrove soils. Insets contain soil carbon (δ13C) and nitrogen 797 (δ15N) stable isotopes with cumulative mass in a low-DBD intact and a deforested mangrove 798 soil. The line at 14 g cm-2 indicates the separation between the upper and bottom reference 799 soil mass layers. 800 801 Figure 5. Excess 210Pb specific activity profiles with cumulative mass in intact (a) and 802 deforested mangrove soils (b). The filled area illustrates excess 210Pb inventories. 803 804 Figure 6. Total ecosystem C stocks of intact and deforested mangrove forests of Tsimipaika 805 Bay, Madagascar. C stocks to 1 m have been corrected for shortening during coring as 806 described in the Methods and should be considered apparent C stocks. 37 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 807 808 Figure 1 809 810 811 Figure 2 38 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 812 813 Figure 3 39 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 814 815 Figure 4 40 This is a post-peer-review, pre-copyedit version of an article published in Ecosystems Journal. The final authenticated version is available online at: https://doi.org/10.1007/s10021-020-00500-z 816 817 Figure 5 818 819 820 Figure 6 41