Assessment of aquatic food web and trophic niche as a measurement of recovery function in restored mangroves in the Southern Gulf of Mexico

Ethical statement

The care and use of animals complied with the Comisión Nacional de Acuacultura y Pesca (CONAPESCA) guidelines and policies. Sampling was carried out with the permit number DF00000156-C from CONAPESCA.

Study area

Located in the southern Gulf of Mexico in the state of Campeche, Mexico, Términos Lagoon is formed by a barrier island and a coastal lagoon system (Fig. 1). It encompasses 120,000 hectares of mangrove, as well as priority sites of biological relevance requiring ecological restoration (Rodríguez-Zúñiga et al., 2013). Characterized by high productivity, biological diversity, and important fisheries, it is a Federal natural protected area for flora and fauna (Yáñez Arancibia et al., 2013). Seawater constantly flows into the lagoon due to mixed diurnal tides with an average amplitude of 0.43 m (range 0.3 to 0.7 m) (Yáñez Arancibia & Day, 2005). There are three climatic seasons in the area. The “nortes” season from October to February is characterized by intermittent rains from winter storms and winds >8 m s⁻¹. The dry season from March to May is characterized by minimal or no precipitation. The rainy season occurs from June to September, with rainfall of 180 mm per month (Yáñez Arancibia & Day, 1982; Yáñez Arancibia & Day, 2005).

Sampling was done on the inland coast of Carmen Island in the Bahamitas Estuary, in four areas with tidal mangrove channels: one reference and three restored (Fig. 1). The latter three channels were dredged by tracing the base preferential flows, which were generated by microwatershed modeling analysis with microtopography information (Pérez-Ceballos et al., 2020). Restoration allowed the recovery of hydrological connectivity with Términos Lagoon, improved the biogeochemical quality, and favored the natural regeneration of the mangroves (Zaldívar-Jiménez et al., 2017).

We selected one area with natural mangroves without apparent modifications, known as reference mangrove (RefM) (18.7028 N, −91.6424 W), where vegetation was dominated by red mangrove (Rhizophora mangle) along the channel edges, with white mangrove (Laguncularia racemosa) and black mangrove (Avicennia germinans) trees and shrubs in the interior. We selected three areas with different restoration times; the area with the longest time since restoration is called RM1 (18.6721 N, −91.6721 W), which was restored in 2010–2011. It is 320 m long and harbors red, white, and black mangrove, although red mangrove dominates. The second area restored is RM2 (18.6928 N, −91.6411 W), which was restored in 2014. It is 1,000 m long, and harbors all three mangrove species, although black mangrove shrubs and red mangrove juveniles dominate the area. The third area restored is RM3 (18.6896 N, −91.6596 W), which was restored in 2018. It is 400 m long and contains all three mangrove species, with red mangrove seedlings and juveniles dominating along with some black mangrove shrubs. In the inner littoral of Carmen Island, adjacent to the study area there are seagrass beds, which constitute an important habitat for species in Términos Lagoon (Yáñez Arancibia, Lara-Domínguez & Day Jr, 1993).

Sampling and laboratory processing

Sampling was done during all three regional climatic seasons: rainy (September 2020); “nortes” (January 2021); and dry (April 2021). Three sampling sites were established in each area to measure environmental parameters. Depth was measured with a graduated ruler. Temperature (°C), dissolved oxygen (DO; mg/L), salinity (ups), total dissolved solids concentration (TDS; mg/L), conductivity (mS/cm), and pH were measured with a YSI Pro multiparameter device. Three water samples were collected in each area to quantify chlorophyll a concentration, which was analyzed with the ethanol extraction method and measured with a spectrophotometer (Nusch, 1980).

Samples of available basal resources were collected. Observed C₃ plants in the restored areas included mangrove (R. mangle, L. racemosa, A. germinans) and saltwort (Batis maritima). Three sets of leaf samples of each plant species were collected and preserved in salt (Arrington & Winemiller, 2002). Seston was separated by filtering 150 ml water with a previously burned Whatman GF/F filter (0.7 µm). Phytoplankton was collected using a cascade filter consisting of 50, 30, and 15 µm sieves. Large volumes of water were run through the filters until the 15 µm sieve was saturated (Sepúlveda-Lozada et al., 2015). The retained fraction was washed with distilled water and filtered through a previously burned Whatman GF/F filter (0.7 µm).

Fish were caught using different fishing gear. Ten baited minnow traps with fish pellets (57.9 cm in diameter × 22.1 cm in height, 0.5 cm mesh) were placed at the channel banks. Five baited cylindrical traps (60 cm long × 26 cm diameter; one cm mesh) were placed in the middle of the channel. A multi-mesh gillnet (55 m long × 2.5 high; 1-to−3.5-inch mesh) was installed at the channel entrance and a cast net (2 m diam., one cm mesh) used when conditions allowed. Fauna sampling was done in the morning at low tide, for approximately 3 to 4 h. Macroinvertebrates (mollusks and crustaceans) were collected with nets or manually.

Collected specimens were preserved on ice for transport to the laboratory. Muscle tissue samples were extracted from the faunal samples of sufficient size: from the muscular foot in the case of mollusks; from the claws in crabs; and from dorsal muscle tissue in fish. For small fish such as poecilids, a sample composed of 5 to 8 organisms was used (Garcia et al., 2007). Tissue samples were preserved in salt and frozen (Arrington & Winemiller, 2002). Specialized keys were used for taxonomic identification of macroinvertebrates (mollusks and crustaceans) (Leal, 2002; Tavares, 2002), and fish (Castro-Aguirre, Espinosa-Pérez & Schmitter-Soto, 1999; Carpenter Kent, 2002; Miller, 2009).

In the laboratory, the plant and muscle tissue samples were washed with distilled water to remove excess salt. They were oven-dried at 60 °C for 48 h and macerated with a mortar and pestle. Subsamples were weighed out (1.5 to 2.5 mg) and placed in tin capsules. These subsamples were sent to the Isotope Laboratory at the University of California, Davis for analysis of stable carbon (¹³C/¹²C) and nitrogen (¹⁵N/¹⁴N) isotopes. Isotopic values were expressed in parts per thousand (‰) relative to standard values in delta (δ) notation. The value is calculated using the following formula: ${\displaystyle \delta X=\left[\left(\text{R sample}/\text{R standard}\right)\right]-1\times 10^3}$ where X is ¹⁵N or ¹³C, and R is the ¹³C/¹²C or ¹⁵N/¹⁴N ratio, using Pee Dee Belemnite as the standard value for carbon, and atmospheric nitrogen for nitrogen. Analytical precision was ± 0.04 for both measurements δ¹³C and δ¹⁵N. Mean SD for the reference materials was ± 0.07‰ for δ¹³C values, and ± 0.05‰ for δ¹⁵N values. The reference materials used for each isotope are listed in (Supplementary Material S1, T1).

Data analysis

Environmental patterns in the tidal channels were elucidated with a principal components analysis (PCA) incorporating the environmental variables documented in the field, with the values log-transformed (x+1) (Supplementary Material S1, T2). The analysis was done with the R ver. 3.6.3 software program (R Core Team, 2021), using the correlation matrix in the vegan package (Oksanen et al., 2019). Interpretation of the PCA was based on those variables with loadings greater than 0.6 (Sedeño Díaz & López-López, 2007).

Identification of the differences in environmental variables between seasons and areas was done with a non-parametric Kruskal-Wallis analysis. This was done because the data did not meet normality and homoscedasticity assumptions. When significant differences were detected, a multiple comparison was made with Dunn’s test with Bonferroni’s correction, with the dunn.test package (Dinno, 2015).

For the basal resources, the δ¹³C and δ¹⁵N values (Supplementary Material S1, T3) were compared between seasons and areas with a non-parametric Kruskal-Wallis analysis and a multiple comparison including Dunn’s test with Bonferroni’s correction. The δ¹³C values of consumers tissues (Supplementary Material S1, T4) were corrected for lipid content according to the mathematical equation provided by Post et al. (2007). Trophic structure at each area was graphed as a biplot using the δ¹³C and δ¹⁵N values for fish, macroinvertebrates and basal resources, including periphyton and seagrass as reported for Términos Lagoon (Sepúlveda-Lozada et al., 2015). The x-axis represents consumer assimilation by organic carbon source (δ¹³C), and the y-axis represents trophic level (δ¹⁵N) (Peterson & Fry, 1987).

Consumer trophic structure was estimated and compared through assembly metrics (Layman et al., 2007). The δ¹⁵N range (NR) provides data on a metric that represents the vertical trophic structure of the community. The δ¹³C range (CR) provides an estimate of the basal resource diversity utilized by the community. Total convex hull area (TA) measures the total amount of isotopic niche space in the community. Mean distance to the centroid (CD) represents the average degree of diversity of isotopic niches and provides data on species spacing. Mean nearest neighbor distance (MNND) measures relative density and species clustering in isotopic space. The standard deviation of nearest neighbor distance (SDNND) measures spatial density uniformity and packing. The isotopic spaces were estimated for each area and season using Bayesian standard ellipse areas (SEA_B). The correction for small sample sizes was applied to estimates of standard ellipse areas (SEA_C). This is a robust method for making statistical comparisons between different groups/samples; in the present case, it allowed comparison of community isotopic niche at different sites and in different seasons. The above metrics were calculated with the SIBER package in R (Jackson et al., 2011).

Basal resource assimilation was estimated with MixSIAR Bayesian mixing models (Stock & Semmens, 2016). The δ¹³C and δ¹⁵N values were corrected by considering the trophic level of the species as in the methods of Abrantes et al. (2013). The trophic level considered was 3.0 for piscivores, 2.7 for zoobenthivores, 2.6 for planktivores and omnivores, and 2.4 for detritivores following Sepúlveda-Lozada et al. (2015). The trophic fraction considered per trophic level was 0.4‰ δ¹³C and 3.4‰ δ¹⁵N following Post (2002). Prior to analysis of the mixing models, Pearson correlations was performed between δ¹³C and δ¹⁵N values of seston and phytoplankton. Because the seston had a significant high correlation with phytoplankton δ¹³C (Pearson’s r = 0.72, p < 0.05) and δ¹⁵N (Pearson’s r = 0.65, p < 0.05), it was not considered in analysis. Also, we verified that consumers were within the mixing polygon defined by basal resources and considering the selected trophic discrimination factors (Supplementary material S2, Fig. 1) (Smith et al., 2013; Phillips et al., 2014). Estimations were made of the contributions of each basal resource to the biomass of all consumers (fish, crustaceans, and mollusks) in each area during each season. Four basal resources were used in the models: C₃ plants, phytoplankton and periphyton and seagrass, reported for Términos Lagoon (Sepúlveda-Lozada et al., 2015) (Supplementary Material S1, T5). The models consider estimated trophic fractionation for carbon (0 ± 1.3‰; (Post, 2002) and nitrogen (0 ± 1‰; Vanderklift & Ponsard, 2003). They were run with a chain length of 300,000 iterations, a burn phase of 200,000 iterations and a thinning of 100. The models produced confidence intervals of each basal resource’s contribution, at a 95% credibility value. They exhibited convergence based on the Gelman–Rubin and Geweke diagnosis (Stock & Semmens, 2016).

The results of the mixing analysis were used to estimate the consumer’s niche n-dimensional hypervolume, to compare the niche between reference and restored mangroves. Each niche dimension represented a basal resource used by consumers (Lesser et al., 2020; Rezek et al., 2020). The mixing models provided an estimate of the basal resources used by each consumer, these data were z-transformed prior to hypervolume construction (Blonder et al., 2014). The hypervolume was estimated using Gaussian kernel density with the hypervolume package in R (Blonder et al., 2018). Overlap hypervolume (Sorensen’s index and confidence intervals) was obtained to estimate similarity between reference mangrove with each restored mangrove areas.

Environmental analysis

The PCA showed that the four analyzed mangroves areas exhibited similar environmental characteristics in response to season. The first two axes explained 68.56% of the variation and all the environmental variables were important (Fig. 2, Supplementary Material S1, T6). We detected environmental differences between seasons. Channel depth was greater in the rainy season (χ² = 24.43, df = 2, p < 0.001), with average values greater than 100 cm (Table 1). Dissolved oxygen (DO) concentrations were low in rainy season (average = 1.1 to 3.1 mg/L, Table 1), with significant differences compared to the “nortes” season (p < 0.001). Values for pH were also low in the rainy season, with values from 7.3 to 7.6 (Table 1, χ² = 16.81, df = 2, p < 0.001). Water temperatures were lowest in the “nortes” season, with values from 22.6 to 30.4 °C (Table 1, χ² = 13.46, df = 2, p = 0.001). Conductivity increased during the dry season, with values greater than 60 mS/cm (Table 1, χ² = 30.30, df = 2, p < 0.001). Salinity values ranged from 39.8 to 48.1 ups (Table 1, χ² = 30.72, df = 2, p < 0.001) and TDS concentrations were higher than 38.8 g/L (Table 1, χ² = 30.30, df = 2, p < 0.001).

Differences between mangroves areas were found for DO (χ² = 11.69, df = 3, p = 0.009) and pH (χ² = 10.08, df = 3, p = 0.02). Dissolved oxygen concentration was higher in RM2 than in RM3 (p = 0.02) and RefM (p = 0.007). Values for pH were also higher in RM2 than in RefM (p = 0.02).

Basal resources

In C₃ plants, δ¹³C average values ranged from −29.3 to −26.4‰ (Supplementary Material S1, T5), with significant differences between seasons (χ² = 14.19, df = 2, p < 0.001) and areas (χ² = 21.15, df = 3, p < 0.001). During the rainy season, C₃ plants had enriched δ¹³C values (p < 0.01). The C₃ plants in RM2 and RM3 were more δ¹³C enriched than in RM1 and RefM (p < 0.01). Averages values for δ¹³C in seston varied between −26.2 and −16.7‰ (Supplementary Material S1, T5), with differences between seasons (χ² = 23.19, df = 2, p < 0.001), ¹³C depleted values in rainy season (p < 0.001), and no significant differences between areas. In phytoplankton, δ¹³C average values ranged from −24.4 to −18.1‰ (Supplementary Material S1, T5), with significant differences between seasons (χ² = 22.82, df = 2, p < 0.001), ¹³C depleted values in the rainy season (p < 0.01), and no differences between areas.

Average values for δ¹⁵N in C₃ plants varied between −2.5 and 3.8‰ (Supplementary Material S1, T5), with significant differences between areas (χ² = 12.59, df = 3, p < 0.001). Differences were observed between RefM and RM1 (p = 0.02), and between RefM and RM3 (p = 0.002). No differences in δ¹⁵N values were observed between seasons. Seston δ¹⁵N average values varied from 1.0 to 3.5‰ (Supplementary Material S1, T5), with significant differences between seasons (χ² = 20.24, df = 2, p < 0.001), and enriched values during the rainy season (p < 0.001). Values for δ¹⁵N in phytoplankton ranged from 2.1 to 4.4‰ (Supplementary Material S1, T5), with differences between seasons (χ² = 12.91, df = 2, p = 0.002), and enriched values during the rainy season (p < 0.01). No differences in δ¹⁵N values were observed between areas in seston and phytoplankton.

Consumers

The consumers samples were composed mainly of fish represented by 22 species, (mostly resident species), mollusks (three species) and crustaceans (five species) (Supplementary Material S1, T4). In consumers, δ¹³C values ranged from −25.3 to −16.6‰ in the rainy season, from −28.2 to −15.0‰ in the “nortes” season and from −26.5 and −12.2‰ in the dry season (Fig. 3). Differences were observed between seasons (χ² = 14.18, df = 2, p < 0.001), with enriched values in the dry season (p < 0.01). Differences were also observed between areas (χ² = 52.66, df = 3, p < 0.001), with ¹³C depleted values in the RefM (p < 0.001).

Values for δ¹⁵N in consumers ranged from 2.0 to 9.6‰ in the rainy season, from 2.8 to 10.8‰ in the “nortes” season, and from 3.7 to 9.9‰ in the dry season (Fig. 3). Differences were observed between seasons (χ² = 19.01, df = 2, p < 0.001), with δ¹⁵N depleted values in the rainy season (p = 0.01). Differences were also observed between areas (χ² = 48.62, df = 3, p < 0.001), with enriched values in RM3 (p = 0.001).

Trophic structure

Consumers in the RefM (fish, mollusks, and crustaceans) used the widest diversity of δ¹³C and δ¹⁵N basal resources (CR, NR, Table 2). This community also exhibited the highest trophic diversity (TA, CD, Table 2, SEA_C Fig. 3) and greatest trophic niche dispersion (MNND, SDNND, Table 2). Among the restored areas, the community of RM1 exhibited the highest values of CR, TA, CD and MNND versus the other restored areas (Fig. 4, Table 2).

Trophic diversity in terms of SEA_C varied between seasons and areas (Fig. 4). The RefM had the highest values in all three seasons, with increases during the dry season. The restored areas had a different pattern, with notably low trophic diversity in RM2, especially during the dry season.

Basal resource contributions

Based on the MixSIAR mixing models, the basal resources contributing to consumer biomass varied between areas and seasons (Table 3). During the rainy season, seagrass and epiphytes constituted the most important resources. The mollusks of all restored areas assimilated seagrass as a main resource, on average from 59% to 77%; C₃ plants were the second most assimilated carbon resource in RM1 and RM3 (19% and 15%, respectively), and epiphytes in RM2 (9%). Crustaceans and fish of all restored areas assimilated mainly epiphytes (from 37% to 72%); phytoplankton was the second most assimilated carbon resource (from 12% to 31%); except for fish of RM1 where seagrass was the second carbon resource (26%). In the reference mangrove, consumers assimilated mainly seagrass (from 40% to 72%) and C₃ plants were the second most assimilated carbon resource (from 15% to 39%).

In the “nortes” season, the contribution of seagrass decreased and that of phytoplankton increased (Table 3). The consumers of RM1 assimilated phytoplankton as main basal resource (57% and 59%), and epiphytes were a secondary carbon resource (18% and 17%). In RM2, the main resources assimilated were seagrass for mollusks (73%); phytoplankton for crustaceans (37%), and epiphytes for fish (55%). The second most assimilated carbon resource was epiphytes for mollusk and crustaceans (12% and 32%, respectively) and phytoplankton for fish (33%). In RM3, the mollusks mainly assimilated seagrass (54%); while crustaceans and fish assimilated mainly phytoplankton (79% and 71%, respectively). The second most assimilated carbon resource were C₃ plants for mollusks (18%), and epiphytes for crustaceans and fish (12% and 23%, respectively). In the reference mangrove, C₃ plants were the main resource sustaining the communities (57% and 53%), the secondary carbon resource was seagrass for crustaceans (20%) and phytoplankton for fish (18%).

In the dry season, almost all consumers in all areas (restored and reference) assimilated mainly phytoplankton, on average from 34% to 88% (Table 3). In RM1, C₃ plants were a secondary carbon source for crustaceans (32%) and epiphytes for fish (32%). In RM2, epiphytes were the secondary carbon resource for the consumers (27% and 9%). In RM3, mollusk assimilated mainly C₃ plants (42%), follow by seagrass (41%); while crustaceans assimilated mainly epiphytes (59%) follow by phytoplankton (33%). In the reference mangrove, C₃ plants were the secondary source assimilated by consumers (32% and 22%).

Trophic niche

The mangrove area with the longest time of restoration (RM1) had a trophic niche hypervolume of 21.3 similar to the reference mangrove of 36.1 (Fig. 5). They had a 49% overlap (Sorensen similarity = 0.49, CI = 0.22–0.5) with similar plants C₃ consumption (Supplementary material S2, Fig. 2). The trophic niche of RM2 was larger (173.1), based on the consumption of epiphytes, phytoplankton, and seagrass. This area had a 19% overlap with the reference mangrove (Sorensen similarity = 0.19, CI = 0.03–0.27). The trophic niche of the area with the most recent restoration (RM3) was 50.2, with consumption similar of RM2, and had a 17% overlap with the reference mangrove (Sorensen similarity = 0.17, CI = 0.01–0.21).