1Sontecomapan is coastal lagoon which, even if relatively small (891-952 ha) (Aké-Castillo et al., 1995; CIMAR, 2011; Contreras, 1985), presents diverse environments due to its geomorphology and a correspondingly complex behaviour. It has a single and somewhat narrow entrance channel at its northeast end, so it can be considered as a choked lagoon (Duck & Da Silva, 2012). It is a brackish system, but its condition varies from marine, during the dry season, to freshwater at the end of its rainy season which result from the interplay between stream input or runoff and tidal influence (López-Portillo et al., 2017).

2As coastal aquatic systems where the exchange between epicontinental water masses and marine water occur, coastal lagoons are sites favouring environmental diversification development which provides habitat opportunities for diverse species, including several with economic importance (De la Lanza & Lozano, 1999). This is dependent on the conditions occurring at these shallow water bodies which can result in high biological productivity at all trophic levels but are, in turn, dependent on the lagoon’s hydrological regime. The hydrological regime is the result of the interaction between the prevailing climatic conditions and the morphometric features of each lagoon, hence providing specific characteristics to each lagoon (González & Nodar, 1981).

3One way to cope with this complexity is to describe coastal lagoons in terms of zones with a similar behaviour instead of doing it on a punctual scale, based on individual sampling sites. Thus, Yáñez-Arancibia (1987) agrees with Fairbridge’s proposal of the following sectors for estuarine environments: a) the low or marine sector, connected to the open sea; b) the medium estuary, submitted to a strong mixing between seawater and freshwater; c) the upper or fluvial estuary, dominated by freshwater but sensitive to daily tidal effects. For Sontecomapan these sectors were recognized by Morán et al. (1995), who reported them as euhaline, polyhaline and mesohaline, respectively.

4Even though, according to the former, the delimitation of these sectors could be performed solely on salinity, evident classification will be best accomplished if other relevant parameters are included by implementing tools such as multivariate analysis. As aforementioned, due to environmental components complexity (i.e., water quality, hydrological parameters, and biological taxa) within the lagoon systems, multivariate statistical approaches have been widely applied to determine the relative influence of environmental factors and their interactions (Kim et al., 2016; Gazzaz et al., 2012; Tobiszewski et al., 2010). This should be done on a temporal and spatial sense, since the extent of lagoon zones is dynamic and will differ from one season to another (De la Lanza et al., 1998; Varona-Cordero & Gutiérrez-Mendieta, 2003).

5The purpose of the present paper is to apply such a multivariate approach to Sontecomapan Lagoon which, due to its geomorphological features, has a complex behaviour and can be subdivided into sectors. This can also establish if there are changes in phytoplankton biomass and trophic status in an interannual scale. In recent years this lagoon has been, at the same time, considered as relatively unpolluted (Gallegos & Botello, 1986; Lot, 1971) but also as a man-disturbed ecosystem (Calva et al., 2005; Ponce-Vélez et al., 1994), so its characterization is important for management purposes, as those areas where water is stagnant would be more sensitive to pollution (Mitchell et al., 2017; Newton & Mudge, 2003). This becomes even more important under the climatic change context.

Materials and Methods

Study area and sampling points

6Sontecomapan is a Mexican coastal lagoon located between 18º30’ N and 18º34’ N and 94º47’ and 95º11’ W, with an area of about 891 ha. Its origin is volcanic, lava flows originally delimited its basin which was further modified by sediment transport through tides, streams and rivers (Lankford, 1977). This lagoon connects to the Gulf of Mexico through a 5.5 m deep, and 137 m wide mouth at its northern end, which then proceeds through a canal towards the Southwest to the main lagoon’s body. The rest of the lagoon is shallower, with a mean depth of 1.5 m and is bordered by mangroves and wetlands which are part of the wildlife protection zone known as Reserva de la Biosfera de Los Tuxtlas (RBLT) (Calva et al., 2005). The lagoon is fed by several rivers and streams, located mainly at its southern and south-eastern portion, though one of its main tributaries, La Palma stream, connects to the canal’s west side (Fig. 1).

7This region’s climate comprises three seasons: north wind season (characterised by strong cold winds, either dry or humid, from the North), dry, and rainy season (Contreras, 1985; López-Portillo et al., 2017). During the rainy season (late June to October), the lagoon receives a continuous freshwater inflow from small rivers and runoff. In contrast, at the peak of the dry season (March to early June), it shows marine salinities. During the Nortes season (November to February), the lagoon displays intermediate conditions, thus being brackish (Aké-Castillo & Vázquez, 2008).

Figure 1. Sontecomapan lagoon. Sampling points. (Modified by Hervé, 2014).

8The annual rainfall ranges between 3000 and 4000 mm. Lagoon water is considered as isothermal, varying from 22 °C in January to 26 °C in May. Generally, water turbidity is high, with a mean Secchi depth of 0.60 m. Water surface temperatures average 24 °C and there is no noticeable temperature gradient down the water column. The southern portion of the lagoon is mesohaline (salinity ranging from 5 to 18 PSU), the central portion is polyhaline (salinity ranging from 25 to 30 PSU) and the mouth is euhaline (salinity ranging from 30 to 40 PSU) (Morán et al., 1995).

9Samples were taken at ten stations (Fig. 1). The sampling station locations (Table 1) were selected as those representative of the diverse ecological and hydrological features; these have been set as interdisciplinary sampling points as in Benitez-Díaz-Mirón et al. (2014).

10Samples were taken in the dry and north wind seasons (March 2009 and January 2010, respectively) and in the rainy season (September 2009 and June 2010), covering the three climatic seasons for this region (López-Portillo et al., 2017).

Physical and chemical water analysis

11Physical and chemical measurements were taken at each sampling station using a Van Dorn sampler 0.3 m below the surface and 0.3 m above the bottom. YSI multiparametric probes were used for temperature and salinity; dissolved oxygen concentrations were analysed using the Winkler method (Strickland & Parsons, 1972). Whatman GF/F glass-fibre filters were used for the spectrophotometric determination of chlorophyll a, through the Jeffrey & Humphrey’s trichromatic method (1975). The filtrate was used to determine the nutrient concentrations (nitrate, nitrite, total ammonia, soluble reactive phosphorus [SRP], and total phosphorus [TP] after persulphate digestion) with a HACH Odyssey DR/2000 spectrophotometer, using the methods described by Strickland & Parsons (1972) and Aminot & Chaussepied (1983). Ammonium (NH₄⁺) concentrations were computed from total ammonia through Emerson et al.’s equation using water temperature and pH (US-EPA, 2013).

Statistical analysis

12Plots of the 95 % confidence intervals for the mean were obtained for each parameter, classifying them by survey and sampling point and by survey and sampling depth (surface or bottom); when these confidence intervals overlap, at least partially, then they are statistically equal. Shapiro-Wilk’s test for normality was applied to data to assess if a parametric ANOVA or a non-parametric Kruskal-Wallis should be applied to each variable. ANOVA was performed to environmental variables to determine if significant variations occurred due to sampling depth, sampling site or survey, and their combinations. Tukey’s HSD (Honest Significant Difference) was applied as a post-hoc test whenever significant differences occurred, for determining which cases were responsible, this was performed on a survey and sampling point basis, and on a survey and sampling level basis. Environmental data (NH4⁺, NO2, NO3, SRP, TP, chlorophyll a, dissolved O₂, and salinity) were transformed to Z (Legendre & Legendre, 1984) for their cluster analysis based on Euclidean distance and Ward’s aggregation algorithm for clustering by sampling sites and survey; values (X) were transformed to log₁₀ (X +1) prior to their ordination through principal component analysis, in order to analyse the links between environmental conditions (Pielou, 1984). For all these analysis, Statistica for Windows 7 (StatSoft, Tulsa, OK) was employed.

Results

Normality tests and ANOVA

13Shapiro-Wilk’s normality test showed that only dissolved oxygen had a normal distribution (Table 2). Thus, the Kruskal-Wallis ANOVA was applied to the rest of the parameters and found significant differences due to survey, and in the case of total phosphorus also by sampling site (Table 3). For dissolved oxygen, the parametric ANOVA found differences due to survey, to sampling point and also to interaction between survey and sampling point (Table 4). None of them found differences due to sampling level (surface/bottom), proving that the water column was mixed and not stratified at any of the sampling points or surveys, as confirmed in Table 6.

Table 2. Shapiro-Wilk’s test for normality

Table 3. Non-parametric Kruskal-Wallis ANOVA

Table 4. Parametric ANOVA for dissolved oxygen

Temporal and spatial distribution of parameters

14Figure 2 presents the 95 % confidence intervals for the mean of each of the water parameters classified by survey and sampling point; Figure 3 classifies them by survey and sampling level. In general terms, mismatch between confidence intervals points to a significant difference; the data responsible for significant differences are shown in the tables 5 and 6, according to Tukey’s post-hoc analysis, the first column contains the differing data, and the second column shows to which other data they exhibits a pairwise difference. Temperature did not differ between surface and bottom, the highest temperatures occurred during June 2010, when temperature was higher at most of the sampling points, attaining 33 °C at Sábalo, except at Levisa and El Real, where the highest temperatures were during September 2009. For salinity (Fig. 2a, Fig. 3a) the graphs show that the minima were found during September (rainy season), when even at La Boya salinity was less than 5 PSU, on the other hand, the highest salinities were in March (dry season), when salinities from 30 to 35 PSU occurred at La Boya, El Chancarral and El Real. Tukey’s test showed that differences also occurred among the sampling point at Basura and the points at La Barra and Chancarral, during the dry season (Table 5).

Figure 2A. 95 % intervals for the mean by sampling point and survey for: A) Salinity B) Dissolved oxygen C) Ammonium D) Nitrite.

Figure 2B. (continuation) 95 % intervals for the mean by sampling point and survey for: E) Nitrate F) SRP G) TP H) Chlorophyll a. BA: Basura, CH: Chancarral, CO: Cocal, CN: Costa Norte, ER: El Real, FR: Fraile, LB: La Barra, LE: Levisa, LP: La Palma, SA: Sábalo.

Figure 3A. 95 %-intervals for the mean by sampling level and survey for: A) Salinity B) Dissolved oxygen C) Ammonium D) Nitrite

Figure 3B. (continuation) 95 % intervals for the mean by sampling level and survey for: E) Nitrate F) SRP G) TP H) Chlorophyll a.

Table 7. Ranges of values for water parameters found at Sontecomapan lagoon.

15For dissolved oxygen, the strongest effect was for the interaction among survey and sampling point with El Cocal, Levisa and El Real, having the highest values during June (rainy season) (Fig. 2b, Fig. 3b). These same sampling points showed the highest chlorophyll a concentrations during the same survey, especially Levisa (Fig. 2h). Ammonium was at its highest during the rainy season surveys of September and June, particularly at El Chancarral and El Real (Fig. 2c, Fig. 3c). The highest Nitrite-N also occurred during the rainy season surveys, especially at La Palma in September, and La Barra in June (Fig. 2d, Fig. 3d). For Nitrate-N the highest concentrations occurred in September at La Palma, and a lesser but still high value at El Fraile, bottom concentrations were higher during this survey (Fig. 2e, Fig. 3e).

16SRP presented the highest values during June at La Palma, El Cocal, Levisa and El Sábalo, at the middle portion of the lagoon (Fig. 2f, Fig. 3f, Table 5). TP was highest in surface samples from the January survey, at El Cocal and Levisa (Fig. 2g, Fig. 3g). In June chlorophyll a presented a value close to 14 µg L^-1 at Levisa and other remarkable values ranging from 3 to 6 µg L^-1 at Costa Norte, El Cocal, Chancarral and La Palma (Fig. 2h).

Multivariate Analysis

17The dendrogram for sampling points at each survey is presented in Figure 4, which represents the average trend for the period described here. Four clusters appeared at an Euclidean distance of 10, which roughly corresponded to the four surveys. The greatest difference took place in June 2010 (except for st. 8), at the onset of rainy season, when the highest chlorophyll a and SRP concentrations were found. The second cluster was mostly for the the rainy season’s peak (September, 2009). A third cluster was for the dry season (March, 2009), but only for the sampling points at the outer and middle portions. These same sampling points, but from the north winds season, formed a fourth cluster’s sub-cluster; at the other sub-cluster the inner lagoon’s sampling points intermingled for the north wind and dry seasons (Fig. 4).

Figure 4. Dendrogram of the sampling points and surveys based on the z-values of the physical and chemical parameters

18Ordination by principal component analysis (PCA) from water variables showed high salinity during the north wind season (January) and the early rainy season (June) at the negative part of the main axis. At the positive extreme ammonium from the dry season (March) appeared, TP from the dry and early rainy season and SRP from the north wind season (Fig. 5). Ammonia from the other surveys appeared at the negative part of the second axis along with oxygen at the early rainy season, SRP at the full rainy season, nitrate at the dry season and nitrite at the early rainy season. The positive part of this second axis relates to dissolved oxygen during the north wind season survey (January), chlorophyll at the full rainy season, TP at the north wind season and SRP at the early rainy season (Fig. 5).

Figure 5. Principal component analysis. Biplot for water variables and sampling points. Sampling points are indicated with a small diamond. The suffix indicates surface (S) or bottom (B). Ba, Basura; Ch, Chancarral; Co, Cocal; CN, Costa Norte, ER, El Real; Fr, Fraile; LB, La Barra; Le, Levisa; LP; La, Palma; Sá, Sábalo. Variables, a small square at the end of a line. The prefix indicates the survey, M (March 2009), S (September 2009), J (January 2010), Ju (June 2010). Chla chlorophyll a, NH4 ammonium, NO2 nitrite, NO3 nitrate, SRP soluble reactive phosphate, TP total phosphorus, Salin salinity

19Sampling points were clustered so that three zones could be distinguished: a mesohaline to oligohaline zone comprising Basura, Fraile and Costa Norte, at the positive part of the first ordination axis. The polyhaline zone comprised Levisa, Sábalo and Cocal, at the positive part of the second axis. The euhaline zone comprised La Barra, Chancarral and El Real, at the negative part of the second axis. La Palma stood as a group on its own, in fact the surface samples clustered to the mesohaline zone while the bottom samples clustered to the euhaline zone; this implies a stratification, contrary to what happened with the rest of the sampling points. This zonation is represented on a map at Figure 6.

Discussion

20During these surveys, for all sampling points the water column in generally was well mixed, as confirmed by the statistical tests applied. In the horizontal axis, three sectors that correspond to those described by Yáñez-Arancibia (1987) were detected, and have also been found for other coastal lagoons, such as the Carretas-Pereyra, Chiapas system (Varona-Cordero & Gutiérrez-Mendieta, 2003), which had been previously described for Sontecomapan by Mondragón (2005) and Morán et al. (1995). Nevertheless, this is the average trend for the period covered by the present research, as there are seasonal variations all along these sectors such as the September survey, when due to heavy rainfall and runoff, salinity lower than 5 PSU was detected at La Boya, the nearest sampling point to sea communication, while euhaline salinities occurred at several points in the March survey. This trend in seasonal shifts along sectors has also been detected in other Mexican coastal lagoons, such as the Chantuto-Panzacola and Carretas-Pereyra systems in Chiapas (Varona-Cordero & Gutiérrez-Mendieta, 2003).

Figure 6. Map of the zones identified through the principal component analysis.

21The stream at La Palma appeared different from the remaining sampling points. At this station, nitrite and nitrate attained their highest values during the September survey, probably arriving with the runoff from human settlements upstream. For SRP, peak values occurred during the June survey; during the rainy season offset, which also points to an upstream origin of SRP. What is remarkable is that phosphorus input was higher at the early rainy season while nitrogen input increased at the season peak. This differs from the trend detected by Aké-Castillo & Vázquez (2008), who applied canonical correspondence analysis (CCA) to their 2002 and 2003 Sontecomapan data and found that salinity positively correlated to SRP, and negatively to silicate, which implies that SRP was of marine origin, while silicate mainly came from rivers and streams; in the present case, SRP entered through stream La Palma as the PCA ordination showed either negative correlation (January) or lack of correlation (June) between SRP and salinity (Fig. 5).

22At coastal lagoons nitrite comes from two sources, either through discharge to water from human activities, or through the oxidation of the ammonia that arises from excretion by ammonotelic organisms or from organic matter decay by ammonia-oxidizing bacteria. Nitrite can then follow one of two paths: it is either further oxidised to nitrate (nitrification) and becomes a nutrient for primary producers, or it enters the reductive path to be transformed into N₂ which is lost from the aquatic environment (denitrification), thus nitrite can be considered as a core chemical species in the nitrogen cycle (Ward, 1996; 2007). The path followed depends on the environment’s oxidative or reductive conditions, and associated microbial assemblages as well. For the present case, it appears that part of the nitrate is derived from human sources and sediment associated to emergent macrophytes and mangrove at the lagoon’s banks, an environment which is fit for ammonia oxidation. In Aké-Castillo & Vázquez’ (2008) CCA ordination, a negative correlation appears between ammonium and nitrate, and ammonium locates close to the mangrove-influenced zones. In López Portillo et al. (2017) the principal component analysis also shows a negative correlation between nitrate or nitrite and ammonium at Sontecomapan. It is to be determined if these results mean that nitrite and nitrate originate from the ammonia oxidation released by organic matter decay at mangrove sediment.

23Mondragón (2005) attempted a trophic characterisation of Sontecomapan by using Vollenwieder & Kerekes’ (1980) probability plots for some parameters, where a relative trophic level is assigned to each value according to the probability for its occurrence, and applied them to two surveys, in February and June 2005. For February, he assigned a hypertrophic category to most of the sampling points, and the rest as eutrophic, based on TP values; while according to chlorophyll a, the sampling points fell into more categories, from oligotrophic to hypertrophic, and were centred in eutrophic. For June, TP values ranged from mesotrophic to hypertrophic, and the same occurred for chlorophyll a. In the present work, the TP values also fall in the range from mesotrophic to hypertrophic, while chlorophyll a ranges from oligotrophic to eutrophic. When comparing these results, the trend for the 2005 values was from eutrophic to hypertrophic, whereas the results reported here ranged from mesotrophic to eutrophic. However, by employing indexes devised for tropical or subtropical reservoirs (Salas & Martino, 1991; Fernandes-Cunha et al., 2013), instead of Vollenwieder & Kerekes’ (1980) index for temperate waters, for the considered period, the corresponding geometric means (0.202 to 0.386 mg L^-1) show that Sontecomapan was hypereutrophic based on TP, while it was oligotrophic for chlorophyll a (geometric means 0.40 to 2.08 mg m^-3). According to Salas & Martino (1991) and Fernandes-Cunha et al. (2013) the corresponding values for the eutrophic onset conditions are 0.078 mg L^-1 for total phosphorus, and 27.2 mg m^-3 for chlorophyll a. This points to the importance of not attempting the trophic characterization of a coastal lagoon on a sole variable basis.

24In fact, a comparison of the chlorophyll a values reported for Sontecomapan since the early 90’s, show that the 2009-2010 data are the lowest for the period (Table 5). Currently, we have no information that would allow us to determine if the system’s productivity had diminished; nutrients were within the ranges reported for the same period, and SRP was at higher concentrations than in former surveys, so that primary production appears not to have been nutrient-limited. Another possible reason for the chlorophyll a and particulate organic matter (POM) diminishing was probably related to exceptional rainfall during this period, especially during the September survey, which probably caused a flushing out of suspended materials to the sea, as confirmed by salinities lesser than 5 PSU near the communication mouth. This agrees with the low zooplankton densities found by Benítez-Díaz-Mirón et al. (2014) in the same period, which the authors attribute to an increased water exchange, among other possible causes. On the other hand, Aké-Castillo & Vázquez (2008) and Muciño et al. (2011) performed surveys in 2002-2003 and 1999, respectively, and detected a phytoplankton flow in both ways, which depended on the tide and rainfall combined effect. Even if flushing protects coastal environments from pollution (Mitchell et al., 2017; Newton & Mudge, 2003), whenever it becomes too intense, it might result in a loss of planktonic production because of its exportation to the sea.

25Aké-Castillo & Vázquez (2008) found that folin phenol active substances (FPAS) released by mangrove leaf litter were positively correlated to the abundance of bloom-forming phytoplankton species, and that this occurred mainly in those zones with high mangrove influence during the dry season, correlating with high salinities. Our PCA ordination presents a contrasting situation, with lower salinities occurring at these sites; taking chlorophyll a as a phytoplankton density proxy, the highest densities were found at the rainy season onset (June) along the sampling points at the channel dividing line from the main lagoon body.

26The point that Aké-Castillo & Vázquez (2008) remark the most from their results is that dominant phytoplankton species at Sontecomapan rank along a gradient stablished according to FPAS, as determined through CCA. This is important, as some of these are harmful algal bloom phytoplankton species (HAB) (Muciño et al., 2011) and FPAS concentration values could be employed to predict the probabilities of HAB occurrence.

27Concerning the use of multivariate methods for the characterization of aquatic ecosystems, they have been successfully employed, in some recent works, for the analysis of lotic, lentic, freshwater and coastal ecosystems, for the identification of spatial or seasonal patterns in their behaviour or else for the detection and assessment of human-induced perturbations extent. For example, Magyar et al. (2013) applied a multivariate approach to assess spatial changes at Neusiedler See, an Austrian lake. Wang et al. (2013), Duan et al. (2016), and Kim et al. (2016) used multivariate statistical techniques for determining water quality patterns in a river, a lake and a canal, while Wang et al. (2015) employed PCA for the characterization of the extent and sources of metal pollution in coastal areas. Liu et al. (2011) linked multivariate analysis to geostatistical methods while Tobiszewski et al. (2010) combined multivariate analysis with experts’ opinion. Srichandan et al. (2015) employed this approach to the study of phytoplankton assemblages at a coastal lagoon. All these researchers agreed in considering these methods as an aid to the patterns and processes understanding in these systems, as well as a necessary step for the construction of predictive models. As an emerging use of multivariate methods, ordination of coastal lagoons has also been applied for identifying those areas related to the breeding and survival of vectors for diseases such as malaria, dengue and chikungunya (Sheela et al., 2015).

28It is expected that the present characterization of Sontecomapan will serve as an aid for its proper management. Duck and Figuereido da Silva (2012) consider that there has been more of a coastal bodies mismanagement which has degraded rather than enhanced the ecosystem services derived from them. This has often resulted from a short sightedness, but also from insufficient knowledge of lagoon processes. At the moment, this becomes even more important, for it is necessary to forecast the effects of an eventual climate change.

Acknowledgements

29The present research was supported by the ECOS-ANUIES-CONACYT M10-A01 project, as part of the CONACYT’s joint program Mexico-France, and the Evaluación diagnóstica de la estructura del circuito microbiano planctónico y de la dinámica hídrica de la laguna de Sontecomapan, Veracruz project (acuerdos del Rector General 12/2008) Universidad Autónoma Metropolitana.

30We acknowledge the comments made to this paper’s draft by anonymous reviewers.