Mangroves provide blue carbon ecological value at a low freshwater cost

At least 11 coastal ecosystems have been considered based on a minimum of actionably defined criteria to be “blue carbon ecosystems”. These include mangrove wetlands, tidal marshes (salt, brackish, fresh), seagrasses, salt flats, freshwater (upper estuarine) tidal forests, macroalgae, phytoplankton, coral reef, marine fauna (fish), oyster reefs, and mud flats⁵; all but three of these would be considered wetlands, with salt flats and mud flats being examples of non-emergent (plant) blue carbon wetlands. Herein, we focus on mangroves.

Adjusting intrinsic leaf-level photosynthetic water use efficiency (({WUE}_{int})) in response to environmental gradients (Introduction)

We used data provided by B.F. Clough & R.G. Sims²⁰, which presented leaf-scale net photosynthesis (({P}_{n}) [sic]; μmol CO₂ m⁻² s⁻¹), stomatal conductance (({g}_{w}): mol m⁻² s⁻¹), leaf-intercellular CO₂ concentrations (({c}_{i}): μl l⁻¹), and intrinsic photosynthetic water use efficiency (({WUE}_{int}): (frac{{P}_{n}}{{g}_{w}})) for 19 mangrove species occupying 9 different sites in Papua New Guinea and northern Australia. These field data were collected using an infrared gas analyzer (model Li-6000, Li-Cor Biosciences, Inc., Lincoln, NE, USA) attached to leaves at saturation light levels (reported as > 800 μmol PPFD m⁻² s⁻¹). Soil salinity at the time of data collection ranged from 10 to 49 psu, and median long-term atmospheric temperature and relative humidity among sites ranged from 19.9 to 27.4 °C and 35.1 to 92.2%, respectively (Fig. S1)²⁰. These data were among the first to offer insight from field study into the plasticity of mangroves across a range of natural salinity and aridity gradients to adjust leaf-level ({WUE}_{int}) as needed for local environmental condition. While it is not new for trees to adjust their ({WUE}_{int}) when they develop in arid, semi-arid, or even some humid and tropical environments⁶⁰, what is distinctive is that mangroves may be further driven to water savings by salinity gradients as a condition of development.

({{varvec{W}}{varvec{U}}{varvec{E}}}_{{varvec{i}}{varvec{n}}{varvec{t}}}) and individual tree water use of mangrove wetlands versus terrestrial ecosystems

For Fig. 1a, we compare leaf-level ({WUE}_{int}) data collected from 17 published papers (using maximum and minimum values), providing 67 independent measurements of ({WUE}_{int}) for mangroves (Table S3). While we mention in the main text that as many as 214 independent measurements of water use efficiency are available, not all of these present raw ({P}_{n}) or ({g}_{w}) data, with some reporting leaf transpiration (({T}_{r})) which do not enable reporting of intrinsic water use efficiencies. Also, we strategically included studies from reproducible experimental designs and readily available papers. Mangrove species included in this review represented a global distribution of greenhouse and field observations, and encompassed species in the following mangrove genera: Rhizophora, Avicennia, Laguncularia, Bruguiera, Aegialitis, Aegiceras, Ceriops, Sonneratia, Kandelia, Excoecaria, Heritiera, Xylocarpus, and Conocarpus.

We then accessed an existing database (n = 11,328 observations) that reported raw ({P}_{n}) and ({g}_{w}) data from 210 upland deciduous and evergreen shrubs and trees of savannah, boreal, temperate, and tropical habitats⁶⁰. From these data, we evaluated a range of upland tree and shrub species that occurred and developed naturally in environments along a global gradient of vapor pressure deficit (i.e., atmospheric moisture and temperature), including arid, semi-arid, dry semi-humid, and humid locations.

For Fig. 1b, we started with a review by Wullschleger et al.⁶¹ that provides maximum individual tree water use data (L H₂O day⁻¹) from 52 published studies representing 67 species of upland trees from around the world. Of those studies, dbh values (8 to 134 cm) were provided alongside 47 individual tree water use values. Maximum individual tree water use and dbh (4.1 to 45.3 cm) were available from the original source for 8 mangrove studies representing 7 species from French Guiana, Mayotte Island (Indian Ocean), China, Florida (USA), and Louisiana (USA) (Table S4). These represent the extent of published sap flow data that provided both individual tree water use and dbh from mangroves (numerically); e.g., we could not extract specific individual tree water use versus dbh from a Moreton Bay (Australia) study site⁶², south Florida study site⁶³, or from five additional study sites in China^51,52. However, regressions for two of the Chinese study sites provided over two years⁵¹ indicated that mangrove trees from a suite of species ranging in dbh from 8 to 24 cm used approximately 0.76 and 9.31 L H₂O day⁻¹, or 0.53 L H₂O day⁻¹ cm⁻¹ of dbh. These apparent rates were even lower than what was reported as average for mangroves in Fig. 1b of 1.4 L H₂O day⁻¹ cm⁻¹. The mangrove species included in this analysis were Avicennia germinans (L.) L., Laguncularia racemosa (L.) C.F. Gaertn., Rhizophora mangle L., Ceriops tagal (Perr.) C.B. Rob., Rhizophora mucronata Lam., Sonneratia apetala Buch.-Ham, and Sonneratia caseolaris (L.) Engl.. Additional comparative mangrove species reported by B. Leng & K.-F. Cao⁵¹ included Bruguiera sexangula (Lour.) Poir., Bruguiera sexangula var. rhynchopetala W.C. Ko, Excoecaria agallocha L., Rhizophora apiculata Blume, Sonneratia alba Sm., and Xylocarpus granatum J. Koenig.

Estimation of canopy transpiration (({{varvec{E}}}_{{varvec{c}}})) from net primary productivity data

Estimates of carbon uptake from CO₂ can provide insight into the water use requirement for that uptake of carbon⁶⁴. We used leaf-level instantaneous water use efficiency (({WUE}_{ins}): (frac{{P}_{N}}{{T}_{r}})), which relates to net CO₂ uptake from leaves of the dominant mangrove forest canopy relative to the specific amount of water used, and developed a predictive relationship (predicted) for determining mangrove net primary productivity (NPP) values from ({E}_{c}) using ({WUE}_{ins}). For A. germinans, L. racemosa, and R. mangle forest components, we used light-saturated, leaf-level ({WUE}_{ins}) values of 3.82 ± 0.3, 4.57 ± 0.3, and 5.15 ± 0.4 mmol CO₂ (mol H₂O)⁻¹ [± 1 SE], respectively, from mangrove saplings and small trees of south Florida⁶⁵. ({WUE}_{ins}) values were stratified by species relative to basal area distributions on each south Florida study plot, converted from molar fractions of H₂0 (from ({E}_{c}) determination) and CO₂ to molecular weights, and multiplied by ({WUE}_{ins}) with applicable unit conversions to attain kg CO₂ m⁻² year⁻¹. This value was multiplied by 0.273 to yield kg C m⁻² year⁻¹.

This predictive relationship was validated in two independent ways. First, for one of the calibration sites (lower Shark River, Everglades National Park, Florida, USA), we modeled ({E}_{c}) from sap flow data⁵⁰, determined NPP from ({WUE}_{ins}) calculations relative to the amount of water the stand used, and had independent measurements of net ecosystem exchange (NEE) of CO₂ between the mangrove ecosystem and atmosphere from an eddy flux tower⁶⁶. For this site, NPP estimation and NEE were closely aligned once soil CO₂ effluxes were accounted; respiratory CO₂ effluxes from soil and pneumatophores were determined to be 1.2 kg C m⁻² year⁻¹ from previous study⁶⁷. Using our NPP estimations from ({WUE}_{ins}) calculations and subtracting soil and pneumatophore CO₂ effluxes of 1.2 kg C m⁻² for 2004 and 0.8 kg C m⁻² for 2005 (partial year), NPP becomes 0.96 kg C m⁻² for 2004 and 0.85 kg C m⁻² for January to August of 2005 (see Observed 1, Florida on Fig. S2). Our approach underestimated NPP from ({E}_{c}) relative to measurements from eddy covariance by 0.21 kg C m^-2 for 2004 (within 17.5% of predicted) and was nearly identical for 2005 (within 0.02 kg C m⁻², or 2% of predicted).

Second, we wanted to determine whether ({E}_{c})-to-NPP predictions developed on a few sites in south Florida, USA, represented other global sites, so we included an analysis from several mangrove sites in Guangdong Province, China, to represent an entirely different location. Similar to south Florida analyses, we combined data for NPP from co-located sites of ({E}_{c}) determination using sap flow techniques. NPP of the mangrove forests were measured using multiple procedures (including eddy flux) for improved accuracy^68,69. The relationships of predicted NPP versus ({E}_{c}) and observed NPP versus ({E}_{c}) did not differ between Florida and China (t = 0.48, p = 0.643).

Projecting mangrove ({{varvec{E}}}_{{varvec{c}}}) to other locations

We reviewed data from 26 published records that report mangrove NPP, or enough data to estimate NPP, from 71 study sites located in the Florida-Caribbean Region (25 sites) and Asia–Pacific Region (46 sites) (Table S5). Table S1 reveals mangrove literature sources used, as well as how NPP was estimated from values provided in the original source; itemizes assumptions for determinations of aboveground NPP, wood production, litter production, and root production from various ratios⁷⁰; and reveals unit conversions.

We then convert NPP to ({E}_{c}) for all 71 sites using the predicted curve in Fig. S2 (Eq. 1, main text), and provide summary results by location in Table S1. Regional (ET) data were extracted from the MODIS Global Evapotranspiration Project (MOD16-A3), which are provided at a resolution of 1-km. The locations of mangrove NPP study sites were identified, assigned to a single 1-km² grid in MOD16, and (ET) was extracted from that grid and used for ({E}_{c})-to-(ET) comparison. Average (ET) from single cells (1 km²) was combined with the average of up to 8 additional neighboring cells to provide comparative (ET) projections over up to 9 km² for each location from 2000 to 2013 to compare sensitivity among suites of the specific MODIS16-A3 cells selected over land. When neighboring cells were completely over water, they were excluded since component mangrove forest ({E}_{c}) estimation was not possible from the cells. Estimates of (ET) by individual cells used to compare with mangrove ({E}_{c}) versus up to 9 cells differed by an average of only 43 mm H₂O year⁻¹ (± 16 mm H₂O year⁻¹, S.E.). Therefore, we use (ET) from individual, overlapping ({E}_{c}) cells in Table S1.

The average ({E}_{c})-to-(ET) ratio from mangroves was subtracted from ({E}_{c})-to-(ET) ratio for specific ecoregions⁴⁸, and this ratio difference was assumed to represent net water use strategy affecting differences by the dominant vegetation between ecosystem types. We were also mindful that salinity reductions can affect ({E}_{c}). We used scaled (0–1) mean and standard deviations from ({WUE}_{int}) data previously reported for mangroves (Fig. S1)²⁰. Standard deviation was 32% of mean ({WUE}_{int}) related to salinity gradients, and if we re-scale this deviation to ({E}_{c}) data and add it to the mean ({E}_{c}) to assume low salinity, average ({E}_{c})-to-(ET) ratio becomes 57.4%. This is theoretical and assumes a relatively linear relationship between ({WUE}_{int}) and ({E}_{c}).

Comparative water use scaling among ecoregions

Table 1 presents the projected reduction in water used through ({E}_{c}) if a mangrove ({E}_{c})-to-(ET) ratio was applied to tropical rainforest (290.52 mm H₂O year⁻¹), temperate deciduous forest (131.76 mm H₂O year⁻¹), tropical grassland (110.77 mm H₂O year⁻¹), temperate grassland (46.48 mm H₂O year⁻¹), temperate coniferous forest (54.96 mm H₂O year⁻¹), desert (22.99 mm H₂O year⁻¹), and Mediterranean shrubland (12.08 mm H₂O year⁻¹). To convert potential water use differences to kL H₂O ha⁻¹ year⁻¹ (as presented in the abstract), the following calculation is used (using the example of tropical rainforest):

For comparisons made to mature (> 12 years) oil palm (Elaeis guineensis Jacq.) plantations, ({E}_{c})-to-(ET) ratio was assumed to range from 53³² to 70%³³, for a water use difference of 1170 and 3160 kL H₂O ha⁻¹ year⁻¹, respectively, relative to annual global mangrove (ET) (of 1172 mm). We multiply these values by the 18,467 ha of land area that was converted from mangroves to oil palm³¹ to attain potential water use differences of 21.6 to 58.4 GL H₂O year⁻¹ from avoided conversion of mangrove to oil palm in this region.

Global water use scaling

In order to determine how much global mangrove area is adjacent to each ecoregion, we conducted a cross-walk between terrestrial ecoregions⁷¹ and those used by Global Mangrove Watch in the 2010 classification of global mangrove area⁷². Terrestrial ecoregions used by Schlesinger & Jasechko⁴⁸ were then able to be associated with specific mangrove areas (Table S6). In other words, given a specific ecoregion, we determined how much mangrove area would be occurring within that same ecoregional geography. Global mangrove area assignment to those ecoregions mapped within 0.1% of the total mangrove area of 13,760,000 ha reported in Bunting et al.⁷². To convert kL H₂O ha⁻¹ year⁻¹ to GL H₂O year⁻¹ among ecoregions, the following calculation was used (continuing with the example of tropical rainforest, which has an area of adjacent mangroves of 112,331.9 km²):

Agent-based modelling of individual tree water use (Discussion)

The BETTINA model simulates the growth of mangrove trees as a response to above- and below-ground resources, i.e. light and water⁴¹. In the model, an individual tree is described by four geometric measures, including stem radius, stem height, crown radius and root radius; attributing functional relevance in terms of resource uptake. Aiming to maximize resource uptake, new biomass is allocated to increase these measures in an optimal but not constant proportion. Water uptake of the tree is driven by the water potential gradient between the soil and the leaves. Thus, porewater salinity is part of what determines the water availability for plants.

With the BETTINA model, we simulated the growth of nine individual mangrove trees under different salinity conditions, ranging between 0 and 80 psu, while all other environmental and tree-specific conditions were kept constant. Simulation time was 200 years so that trees could achieve very close to their maximum possible size, and the hydrological parameters were similar to that reported previously⁴². We can show that the ratio of the actual transpiration to the potential transpiration decreases with increasing salinity; plants use less water. Potential transpiration was the transpiration of a given tree without a simulated reduction in water availability due to porewater salinity. These parameter details are presented graphically for mangroves (Fig. S3), comparing porewater salinity along a gradient against the ratio of actual-to-potential individual tree transpiration.

Further, BETTINA simulation results include morphological plasticity adjustments to allometry. To highlight this, we also displayed results assuming a constant allometry as for 40 psu. Naturally, for this arbitrary benchmark the solid and the dashed line coincide (Fig. 3a). Adaptation to higher salinities improves water uptake (primarily girth and root growth), thus the adapted trees (solid lines) have a higher water uptake than the average allometry (dashed lines) for salinities below 40 psu. Lower salinities promote increase of height and crown radius to improve light availability. That is why the adapted trees have a lower water uptake than an average tree would for salinities above 40 psu. Tree water use decreases with increasing salinity (Fig. 3b), as ({WUE}_{int}) coincidently increases (Fig. S1).

Virtual water use explained (Discussion)

Water is required to produce products or acquire services from natural ecosystems; e.g., forest products, fisheries biomass, nutrient processing (nitrification, denitrification), food production. If a net kilogram of a food is grown on a hectare of land where water is abundant and that kilogram of food requires 400 mm of water to be produced, the export of that food to an area of low water availability provides an ecosystem service in the amount of 1 kg of food, plus 400 mm of “virtual” water not actually needed at the destination but used at the source. This water is defined as the product’s “virtual water content”⁵⁶. There is a rich body of literature exploring the concept of virtual water^73,74, but we expand on this concept here as a comparison among 7 ecoregions⁴⁸ and mangroves. Raw data used for calculations are presented in Table S2.

Statistical analysis

Data for leaf-level ({WUE}_{int}) comparisons between terrestrial woody plants and mangroves, as well as individual tree water use by dbh for both terrestrial and mangrove trees, were not normally distributed. We used a Kruskal–Wallis ANOVA based on ranks, and the Dunn’s Method for difference tests. Individual tree water use by dbh for both terrestrial and mangrove trees were determined using linear regression, mostly applied to mean values. For a couple of mangrove studies, only median values were extractable from minimum and maximum values. Likewise, all other data relationships were best fit with linear models, including the calibration curves between ({E}_{c}) and NPP. All data were analyzed using SigmaPlot (v. 14.0, Systat, Inc., Palo Alto, California, USA).

Source: Ecology - nature.com