The Red Sea ranks amongst the most oligotrophic seas due to lack of river runoff and low nutrient supply (Mandura, 1997), and mangrove trees have been shown to be strongly nutrient limited, particularly by iron, in the Central Red Sea (Almahasheer et al., 2016b). Mangroves in the Central Red Sea develop as dwarf trees (<2 m height) along a narrow belt. The Central Red Sea is the most oligotrophic part of the basin indicative of low Chl a concentrations Sea (Raitsos et al., 2013). And the low nutrient concentration could be due to the absence of riverine inputs and limited exchange with Indian ocean (Thompson et al., 2013). Avicennia marina in the Central Red Sea is subjected to extreme arid condition with salinity above 38 psu and water temperature as high as 35°C (Almahasheer et al., 2016c). This study was conducted in ten trees within four different locations in the Central Red Sea. Mangrove locations were composed of monospecific stands of Avicennia marina, away from human disturbances in Khor Alkharar (22°58′14.81″N 38°50′44.27″E), and close to petrochemical activities in Petro Rabigh (22°45′5.81″N 39° 0′46.90″E), while those at Economic City lagoon (22°23′31.10″N 39°7′49.07″E) are under development. Those at Thuwal Island (22°16′54.42″N 39° 5′6.12″E) are less impacted by human activities (22°16′54.42″N 39° 5′6.12″E, Figure 1). Additional information about the study locations and their environmental conditions are described in (Almahasheer et al., 2016b,c).

Leaves and sediment cores were collected in 2014 and 2015 from each location. A total of two to three long axillary branches in each location were randomly selected, to collect a total of 91 leaves from 10 trees. The collected leaves were gently washed with sea water and numbered in the field based on their location in nodes along the axillary shoot starting from the first leaf near the meristem to the last attached senescing leaf. Sediments were sampled from the same locations using 170 cm long PVC cores, with a total of 26 sediment core (6 cores at Khor Alkharar, 4 cores at Petro Rabigh, 8 cores at Economic City and 8 cores at Thuwal Island). Sediments cores were sliced into 1 cm or 3 cm slices down to (24–27 cm compressed, which is equivalent to 26–45 cm decompressed), depending on the core-type (Almahasheer et al., 2017). The slices were dried in the oven at 60°C to constant weight, after measuring their fresh volume to calculate bulk density. The sediments were then subsampled for chemical analyses (C, N, P, and Fe). Then to allow direct comparisons among locations, nutrient stocks in sediment samples at depth (26–45 cm decompressed) were standardized to 30 cm.

Each leaf sampled was individually photographed, dried at 60°C till constant weight, then each leaf was weighted before grinding with a mortar. We then weighted approximately 2 mg of each powdered leaf to the nearest 0.001 mg with an ultra-micro balance in a pre-combusted aluminum capsule, which was then used to determine the concentration of C and N on a FLASH 2000 CHNS Analyzer (Zimmermann et al., 1997). For C and N analyses in the sediment, 1 g of dried ground samples was acidified with 1 M HCl to remove inorganic carbon, centrifuged (3,500 RPM; 5 min) and the supernatant with acid residues was removed using a pipette, then washed in deionized water once, the residues were centrifuged again and the supernatant removed. The residual samples were re-dried (70°C) and encapsulated for C and N analysis using a Thermo Delta V Conflo III coupled to a Costech 4010 at the UH Hilo Analytical Laboratory, USA. The C and %N was calculated for the bulk (pre-acidified) samples.

For Iron and phosphorus concentrations, 0.50 mg dry weight (DW) of each powdered leaf were digested in 6 ml HNO₃ and 2 ml of H₂O₂ following Spalla et al. (2009), and 200 mg of sediment were digested in 6 ml of HNO₃ and 2 ml of HCL following the EPA method 3052 (Kingston and Walter, 1995). The samples were let to cool then diluted with Mili-Q water to be subsequently analyzed by Inductively Coupled Plasma-Optical Emission Spectrometry (Varian Inc. model 720-ES). Analyses of replicates (two replicates between each 20 sample with total of 10) and reference materials (e.g., BBOT and Sulfanilamide in the leaves, NIST 8704 in the sediment for N and C; and Inorganic Ventures and PerkinElmer's Pure Plus for P and Fe in the leaves and sediment) were carried out to ensure reproducibility of the results. Then, mean (±SE) of nutrient concentration (g DW Kg⁻¹) and accumulated stock per unit area (i.e., cumulative mass; Kg m⁻²) in the top 30 cm of the sediment was estimated from the four locations.

The nutrient content in mg leaf⁻¹ was calculated as the product of the nutrient concentration in mg g DW⁻¹ and the total dry mass in g DW leaf⁻¹ of the leaves Lin and Wang (2001; Figure 2). We also calculated the nutrient resorption after Lin and Wang (2001) and Lovelock et al. (2004), as the ratio between the difference in average nutrient content in mg leaf⁻¹ from a fully developed or mature leaf and that of senescing leaves, free of grazing damage, and reported this as a percent. Leaf age was estimated using the plastochrone interval approach, i.e., the time in between development of a new node supporting a new leaf pair (Erickson and Michelini, 1957; Duarte et al., 1994; Coulter et al., 2001). Avicennia marina produced 9.59 nodes year⁻¹, resulting in an estimated plastochron interval of PI = 38 days (Almahasheer et al., 2016c), the time interval between production of two consecutive leaf pairs along a branch.

To calculate the leaf export flux of nutrients in senescing leaves in mg element m⁻² year⁻¹, we multiplied the mean nutrient content of each element in the senescing leaves with leaves produced m⁻² year⁻¹. We simply estimated number of leaves produced m⁻² year⁻¹ after Coulter et al. (2001), which is the product of the number of meristems m⁻² multiplied by the number of internodes produced per meristem per year multiplied by two which is the number of leaves per meristem (Figure 2). Specifically, (1) we used a line transect of 3 m wide by 5–15 m long in two locations (n = 74 trees for Khor Alkharar and n = 153 tree for Thuwal Island, Table 1) to measure each tree/ seedling girth and number of meristems, then used as regression analysis to get the total number of meristems in the two locations (Figure 3). Then (2) we used the number of internodes year⁻¹ for the same two locations above that were obtained from (Almahasheer et al., 2016c) (9.56 node year⁻¹ for Khor Alkharrar and 9.72 year⁻¹ for Thuwal Island), and multiplied by 2 which is number of leaves per node, following Duke and Pinzon (1992) as the rates of annual node production = half of leaf production.

Statistical analyses, including descriptive statistics, linear regression analyses of leaf age vs. element, general linear models to test differences among stands, and Tukey HSD posthoc test to assess pairwise differences were carried out using JMP, a computer program for statistics developed by the SAS Institute.

The number of meristems in Avicennia marina trees increased with tree girth both in Khor Alkharar and Thuwal Island (R² = 0.95, F = 1350, P < 0.0001; and R² = 0.98, F = 8809 and P < 0.0001, respectively; Figures 3A,B). When combined with tree density, girth distribution and previous estimates of nodes production (Table 1), these relationships allowed to estimate that Avicennia marina in Khor Alkharar produced 7871.5 leaves m⁻² year⁻¹ and the stand in Thuwal Island produced 7345.2 leaves m⁻² year⁻¹, which, provided the life span of Avicennia marina leaves is ≤1 year, is equivalent to the leaf litter fall fluxes, as grazing losses were minimal.

The average leaf nitrogen concentration in mg N g DW⁻¹ tended to be low and was significantly higher in Petro Rabigh compared to the other mangrove stands, where average leaf nitrogen content did not differ significantly among these locations (Table 2A). Similarly, the average leaf phosphorous concentration was also low (Table 2A). Leaf P concentration and content did not vary significantly among stands (Table 2A). Likewise, the average leaf iron concentration was very low (Table 2), and both the average leaf iron concentration and content were higher in Petro Rabigh and Economic City compared to Khor Alkharar, while Thuwal Island showed intermediate values (Table 2A).

The nitrogen and phosphorous concentration in the leaves declined with leaf age as a result of either or both the dilution or leaching of nutrients as the leaves mature and expand (Figure 4). However, iron concentration was stable or tended to increase with leave age (Figure 4). The leaf nutrient content was highly variable between trees, with a general trend to decline in leaves around 200 days old in most of the trees suggesting a retrieval of nutrients from old leaves before these are shed (Figure 5). The trends in elemental concentration and content with leaf age were described, based on regression analysis (Tables S1, S2 and Figures S1, S2). The leaf P: Fe ratio declined sharply with leaf age from 20 in young leaves to around 3 in mature 200 days old then a ratio of 1 in senescing leaves, similarly, N: Fe also declined sharply from values of >1,000 to 100 in mature leaves of 200 days old then 50 in senescing leaves. Contrary, the leaf N: P ratio showed a general trend to increase from values of 30 in young leaves to around 40 at mature leaves of 200 days old and remained the same in senescing leaves (Figure 6).

N and P resorption was detected in all sampled trees, whereas Fe resorption was not detected in one out of the 10 trees (Table 2B). There was no significant difference in resorption efficiency between locations for either N, P, or Fe (Table 2B). Hence, we calculated the grand mean (±SE) of resorption to be 69 ± 4% N, 72 ± 2% P and 42 ± 5% Fe, with Fe having lower resorption efficiency compared to N and P. Resorption efficiency was independent of the maximum nutrient concentration and, therefore, independent of the nutrient status of the plants. However, resorption of N and P were significantly correlated (r_N−P = 0.8, P = 0.004), whereas that of Fe with either N or P was not significant (r_Fe−P = 0.6, P = 0.08; r_N−Fe = 0.4, P = 0.29). Moreover, leaf stoichiometry (1918 C:36 N:1 P: 0.5 Fe for all leaves, Figure 7) showed a shift from mature leaves (1715 C: 39 N:1 P:0.5 Fe) to senescing leaves (4213 C:40 N:1 P: 1.1 Fe) involving a decrease in N/Fe and P/Fe ratios by half during senescence (Figure 7).

Both sediment nutrient concentration in g DW Kg⁻¹ and accumulated nutrient stock in sediment (Kg m⁻²) averaged over the top 30 cm of sediment did not differ significantly among sites for both nitrogen and phosphorous, whereas the iron sediment concentration was significantly higher in Petro Rabigh, followed by Economic City, Khor Alkharar, and mangroves in Thuwal Island showing the lowest Fe sediment stocks (Table 3).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.