The Mar Menor is a choked coastal lagoon located in SE Spain, a semi-arid region of the SW Mediterranean (Figure 1). The mean annual rainfall in the area is less than 300 mm/year, and potential evapotranspiration is close to 900 mm/year (López Bermúdez et al., 1981). So, the net hydric balance attained an annual deficit above 600 mm/year (Pérez-Ruzafa et al., 2005a).

Traditionally, the Mar Menor has been markedly oligotrophic, its main distinctive feature being the transparency of its waters compared to the great majority of coastal lagoons. The primary production was mainly benthic, based on microphytobenthos and on the phanerogam Cymodocea nodosa (Ucria) Ascherson (1870) as the main macrophyte. After the enlarging of one of its inlets (El Estacio) in 1973, benthic vegetation evolved toward a C. nodosa–Caulerpa prolifera (Forsskål) J.V. Lamouroux, 1809 mixed bed, with a biomass of approximately 280 gDW/m² (Pérez-Ruzafa, 1989; Pérez-Ruzafa et al., 1989; Terrados and Ros, 1991). This high benthic biomass contrasted with the low phytoplankton density (Ros and Miracle, 1984) and water oligotrophy (Gilabert, 2001a). It was estimated that 63.18% of the total primary production of the lagoon was due to C. prolifera, 0.42% to C. nodosa, 0.24% to photophilic algae, 11.62% to microphytobenthos, and 24.53% to phytoplankton (Terrados and Ros, 1991).

Despite the existence of more than 20 cataclinal watercourses in its watershed, traditionally, none of them maintained a permanent flow toward the lagoon, conditioned by a sporadic and torrential rainfall regime (Figure 1). However, since the early 1990s, after changes in agricultural practices from dry farming to irrigation and related phreatic rising, the main collector in the drainage basin, the Albujón watercourse, has maintained a regular flux of water (Pérez-Ruzafa and Aragón, 2002). This watercourse is about 40 km long, draining an area of 556 km². The flow of El Albujón was estimated to be around 0.02 m³/s at about 4 km from its mouth, reaching up to 10.5 m³/s during a typical storm event (return period of 5.75 years) (García-Pintado et al., 2007). The load of nutrients that flow down the watercourses toward the lagoon shows some temporal pattern, with the highest contents of nitrates (>200 mg NO3−/L) and ammonium (>30 mg NH4+/L) in periods of maximum agricultural activities and of phosphorus (>10 mg PO43−/L) in summer season (Álvarez-Rogel et al., 2006).

As a result of the progressive increase in nutrient inputs, the trophic status changed from oligotrophic to eutrophic. Consequently, the assemblages dominated by small flagellates (Rhodomonas spp. and Cryptomonas spp.) in winter, and diatoms and dinoflagellates from spring to autumn, were replaced by assemblages constituted by large diatoms (Coscinodiscus spp. and Asterionella spp.) present throughout the year, but still maintaining low phytoplankton biomass and without any significant increase in chlorophyll a concentration (Pérez-Ruzafa et al., 2002). The maintenance of good water quality was due to the proliferation, since 1995, of the jellyfish Rhizostoma pulmo (Macri, 1778) and Cotylorhiza tuberculata (Macri, 1778), whose population was estimated at 40 million individuals, exerting a top-down control of the trophic web (Pérez-Ruzafa et al., 2002). This situation had remained relatively stable for 20 years, when a change in the water quality occurred that became particularly evident, even at social level, in the spring of 2016. The water column became dominated by dense populations of Synechococcus spp., and the depth of visibility in the water column was reduced from around 6 m to less than 0.5 m, whereby the compensation depth was above the average depth of the lagoon (4.5 m), leading to the mortality of the C. prolifera–C. nodosa mixed meadows located at more than 2 m depth (Pérez-Ruzafa et al., 2018).

Spatial and temporal variations of hydrological conditions, nutrients, and chlorophyll a concentration in the Mar Menor lagoon have been analyzed in different projects over the last 22 years, using a sampling stations network that spatially covered the lagoon and following the same protocols (Figure 1). In 1997, weekly surveys were conducted from February to December, while bi-monthly surveys were carried out from May 2002 to May 2003, from February 2006 to September 2013, and from February 2016 to November 2018. According to the terrestrial and marine influence, 20 sampling stations were established in the lagoon, grouped into five zones, so that each zone was represented by four replicate sampling units. Since 2009 eight new sampling stations were added, three in the inner mouth of the lagoon inlets and five in the Mediterranean. More water samples have been also obtained in the framework of the different projects. In total, 5780 water samples have been taken during this period.

Water samples were taken at an approximate depth of 1 m, by pumping or using a Niskin bottle, kept in the dark at 4°C in the field and stored at -28°C. Nitrate (N-NO3−), nitrite (N-NO2−), ammonium (N-NH4+), phosphate (P-PO43−), and silicate (Si-SiO44−) were determined following the methods described by Parsons et al. (1984) and using a continuous flow autoanalyzer (SYSTEA μMAC-1000 and SEAL AutoAnalyzer 3 with a JASCO Fluorescence detector FP-2020 Plus). In 1997, salinity was determined with a Beckman RS 7B salinometer and chlorophyll a was analyzed with the spectrophotometric methods reported by Parsons et al. (1984) in all sampling stations. Since 2002–2003, in situ determinations of salinity, temperature, dissolved oxygen, and chlorophyll a were performed using a WTW Multiline F/Set3 multiple probe, and since 2016 a YSI 6600 multiparameter probe was used including also turbidity measures. Samples for chlorophyll a analysis, following Parsons et al. (1984) spectrophotometric methods, were maintained in nine sampling stations during each campaign for the calibration of probe data. The field campaigns for ichthyoplankton composition studies were carried out using a 500 μm-mesh net equipped with a digital flowmeter (General Oceanics 2030) fixed to the mouth to calculate the volume of filtered water. The methodology is described in detail in Pérez-Ruzafa et al. (2004, 2005a).

Two matrices have been built, one with the data of all the sampling locations in each campaign (complete matrix) and the other averaging all sampling stations in each campaign (campaigns’ matrix). In both cases, the localities of the Mediterranean, outside the lagoon, have been excluded. Meteorological variables (global radiation, rainfall, and wind speed), obtained from the Spanish Meteorological Agency at the San Javier station on the Mar Menor coast, were added to the campaigns’ matrix. Climatic data, obtained with hourly resolution, were averaged per each sampling day and for the week prior to it.

Relationships between the hydrological, meteorological, and trophic variables were tested using Pearson’s correlation. Temporal differences throughout the process of eutrophication have been analyzed using a permutational multivariate analysis of the variance (PERMANOVA) (Anderson, 2001, 2005), considering three factors: Trophic status phase, Year, and Season. Trophic status phase was considered fixed with three levels, Prebreak from January 2010 to June 2012, Break from February 2016 to December 2017, and Recovery from January 2018 to November 2018. Year and Season were considered randomly, nested in Trophic status phase. By using permutations, the test does not require specific assumption concerning the number of variables or the nature of their individual distributions or correlations (Anderson, 2001). A random subset of 9999 permutations was used. Data were previously normalized and the analyses were performed on the campaigns’ matrix, using the Primer 6.1 package. Euclidean distances were used for the similarity matrix. Significant terms were investigated using a posteriori pairwise comparison with the PERMANOVA t-statistic and 9999 permutations. The contribution of each variable to average differences between factor levels was examined using Similarity percentage analysis (SIMPER) (Clarke et al., 2014).

Relationships among the considered variables and patterns during the eutrophication process were analyzed using a principal component analysis (PCA) and a multidimensional scaling analysis (MDS), using the campaigns’ matrix after square root and log(x+1) transformation of the environmental and ichthyoplankton variables, respectively.

The best linear model accounting for the observed variation of chlorophyll a and water transparency was analyzed using multiple linear regression models (GLM) with stepwise forward selection of variables (using P < 0.05 as the inclusion criterion) performed on the complete matrix.

As a standard reference for the evolution of ecological status in the water column, TRIX index was calculated for each sample according to the algorithm of Vollenweider et al. (1998) and compared with Secchi disc transparency data.

where DIN = N-NO3− + N-NO2− + N-NH4+.

The Mar Menor area is characterized by high global irradiation. In the three studied phases (Prebreak, Break, and Recovery), it showed the minima seasonal average of daily global irradiation in autumn, ranging between 3062.51 ± 174.81 and 3806.61 ± 272.18 wats/m², and the maxima in spring and summer ranging between 6387.78 ± 421.33 and 7471.46 ± 291.74 wats/m². On the contrary, rainfall is very low and usually concentrates in short intense rain events. Cumulated rainfall in the week previous to sampling days ranged between 0.18 ± 0.10 L/m² in spring and 39.66 ± 32.81 L/m² in autumn, both in the Break phase (Table 1). However, some sporadic peaks reached 100 or >200 L/m² in 1 week, such as the event that occurred in September 2009 and December 2016 (Figure 2).

Water temperature showed a quite regular seasonal and interannual cycle (Figure 3). The minima seasonal average water temperatures were reached in winter, ranging between 12.78 ± 0.14 and 13.65 ± 0.09°C in the three periods considered, while the maxima were reached in summer, ranging between 27.99 ± 0.04 and 28.4 ± 0.09°C. Salinity showed, however, a very heterogeneous temporal dynamic (Figure 3), ranging between 41.48 ± 0.15 in winter and 45.23 ± 0.16 in autumn (Table 2).

PERMANOVA analyses performed on meteorological and hydrographic (salinity and temperature) data on the campaigns’ matrix showed no significant differences in climatic or hydrographical conditions between the three trophic status phases, but with significant interannual and seasonal variability (Table 3).

SIMPER analyses showed that the main differences between pairs of years were determined by the Maximum wind speed and Rainfall. The latter mainly determining the differences between 2007, 2008, and 2009. At the same time, high rainfall and low salinity determined the differences between 2010, 2011, and 2016 with respect to the remaining years. Although low salinity is usually associated with rainfall, in the case of Mar Menor it can also be determined by the Albujón watercourse wastes coming from irrigation, that are usually more intense in dry periods. Furthermore, the surplus of rain and irrigation recharges the phreatic, producing subsurface discharges in the coastal areas of the lagoon. This takes place with a temporary delay with respect to the rains and can last for long periods of time. That was the case after the strong rainfall occurred in December 2016 that produced a decrease in salinity maintained until May 2017 (Figure 2, 3). In fact, 2017 was characterized by low rainfall but at the same time low salinity during most of the year.

Trophic variables and ichthyoplankton assemblages in the water column showed very irregular patterns with very low correlation with climatic and hydrological parameters and between them (Figure 4, 5). The highest significant correlations were found between dissolved oxygen concentration in mg/L and irradiation and water temperature (-0.42; P = 0.00 and -0.64; P = 0.00, respectively), and between chlorophyll a and silicate (0.40; P = 0.00). Chlorophyll a and ammonium also showed a moderate correlation (0.40) but not a significant one (P = 0.43).

While seasonal average nutrient values in the three trophic phases analyzed showed low-to-moderate concentrations (Figure 4, Table 2), all nutrients had peaks of greater or lesser intensity throughout the study period. Seasonal mean values were lower than 4.23 ± 0.34 μmol/L in the case of nitrate, 0.23 ± 0.01 μmol/L for nitrite, 10.09 ± 0.61 μmol/L for ammonium, and 0.39 ± 0.04 μmol/L for phosphate. The spatial distribution of the maximum values of nitrate and chlorophyll a showed very distinct patterns in the three trophic status phases (Figure 6). Nitrate reached 145.79 μmol/L in some lagoon localities in 2006 and up to 263.19 μmol/L close to the Albujón watercourse, both during the Prebreak period, while maximum values were always lower than 37.5 μmol/L during Break and Recovery phases. On the contrary, Chl a reached its maximum concentration during the Break phase (49.7 μg/L) and its values were lower than 27 and 13.5 μg/L during Prebreak and Recovery phases, respectively.

The remaining nutrients also showed distinct patterns between periods. Nitrites maintained low values. They only exceeded 2 μmol/L during the Prebreak phase, in eight occasions, all of them in November–December 2006 and in spring and summer of 2007, all in coastal localities, mainly in the mouth of watercourses, reaching in three occasions values higher than 5 μmol/L (7.18, 32.97, and 47.61 μmol/L). The highest values of phosphate were reached early during the Prebreak period. In 1997, values between 2.20 and 8.74 μmol/L were found in most sampling localities, from the end of February to the end of April. In July 2002, still the early Prebreak phase, the highest punctual values of phosphate were reported, reaching 23.86 and 27.17 μmol/L. Silicate, on the other hand, reached the highest values during the Break and the Recovery phases with mean values of 35.04 ± 1.01 and 14.81 ± 0.60 μmol/L, respectively. Its highest values being at the end of the Break phase, reaching values frequently higher than 100 and up to 133.61 μmol/L. Ammonium reached values higher than 30 μmol/L only in two occasions, 40.42 and 44.63 μmol/L, in March and July 2007, respectively, close to the mouth of the Albujón watercourse during the Prebreak phase. However, it reached values between 30.35 and 37.25 μmol/L in several occasions in the inner lagoon localities in April, July, and August of 2017, during the Break period.

PERMANOVA analyses performed on the trophic and water quality variables, using the factors Trophic status phase (Ts) as a fixed factor, and Year and Season nested in Ts, showed significant differences between years and for the interaction Year × Season, precluding to detect differences between phases (Table 4). However, these differences became highly significant when the analysis was performed considering only the Trophic status phase factor (Table 5).

MDS analysis performed on the same trophic and water quality variables from the campaigns’ matrix confirmed this result and showed the same samples distribution than PCA in Figure 7. ANOSIM analyses also confirmed significant differences between all pairwise comparisons of Trophic status phases (P < 0.003).

Multiple regression analyses with forward selection of variables showed that there is a low percentage of chlorophyll a concentration variance that is explained by meteorological, hydrological, or nutrient variables. This implies that there are other factors that determine the concentration of chlorophyll a or, more probably, that there are complex patterns and temporary lags in the response of chlorophyll a to them. This analysis, performed both on campaigns’ and complete matrices, showed a similar adjusted squared multiple R of 0.29 and 0.30, respectively (Tables 6, 7). The first case considered the averaged data for the whole lagoon in 151 campaigns carried out over the past 22 years, and the last one considered the 1324 sampling stations sampled in the same 151 campaigns and excluding the meteorological conditions. Although in both analyses NO3−, PO43− and SiO44− showed a positive coefficient, indicating that chlorophyll a tends to increase with these nutrients, the response is not linear (Figure 8) and the concentration of chlorophyll increases mainly at low concentrations, but tends to decrease at higher concentrations of them. Ammonium produced the contrary pattern effect on chlorophyll a concentration. Chlorophyll a decreases when increasing ammonium at low concentrations, but tends to increase when this nutrient reached concentrations higher than 25 μmol/L. In fact, in the case of campaigns’ matrix, dissolved inorganic nitrogen (DIN) showed a negative, but not significant, coefficient, probably due to the influence of NH4+ which in the complete matrix analyses was selected with a significant negative coefficient.

These complex patterns are consistent with trophic web regulations on chlorophyll a and suggests that zooplankton produces ammonium at while grazing on phytoplankton. However, when nitrate is low and the production of ammonium is high during the Recovery phase, the phytoplankton can grow using this last nutrient.

Regarding meteorological and hydrographical conditions affecting chlorophyll a concentration in the water column, only the average wind speed of the day of sampling showed positive influence, while maximum wind speed in the same day showed negative coefficient. This implies that while a moderate and sustained agitation of the water column favors planktonic primary production, wind storms produce the opposite effect. At the scale of localities, only salinity and turbidity showed a significant positive relationship with chlorophyll a. In the case of salinity, this is because the maximum chlorophyll a concentration tends to be higher in the confined areas, especially in the Break phase. Chlorophyll a shows especially low values in the areas of direct influence of the watercourses and El Estacio inlet, in which the salinity is low due to the inputs of fresh and less salty Mediterranean waters, respectively (Figure 6D–F). In the case of turbidity, the positive correlation can be explained by the direct effect of the chlorophyll itself on this parameter.

The first two axes of the PCA explain a cumulative 49.87% of the variance (Figure 7). PC1 axis shows 29.48% and is related to high water transparency and high ichthyoplankton species richness and abundance in the positive part, and high nutrient concentration, mainly SiO44−, and Chl a in the negative one. It clearly separates Break and Prebreak phases. Recovery samples, being partially overlapping Break situation, are displaced toward Prebreak conditions, overlapping them. PC2 axis, showing an additional 20% of the variance, discriminates mainly winter samples, with situations with higher DIN, ammonium, and N/P ratios in the positive part and spring and summer situations with high Chl a and PO43− in the negative one.

The opposite situation of ichthyoplankton richness and abundance versus chlorophyll a concentration along the PC1 axis suggests a top down control of the trophic web during the Prebreak and Recovery phases.

The water column transparency is mainly constrained by Chl a concentration, that explain up to 55% of Secchi disc visibility at a given locality. In fact, chlorophyll a is the only variable selected by the GLM regression model with forward selection of variables performed using 1188 samples of the complete matrix (Figure 9A). This percentage increased up to 57% when considering the campaigns’ matrix.

In summary, Prebreak phase is characterized by high water transparency, the highest peaks of nitrates and phosphate and maximum values of nitrate, but also by the lowest concentrations of silicate, nitrites, ammonium, and chlorophyll a, despite the high nutrient loads in the Mar Menor. Break phase was characterized by the highest average nutrient and chlorophyll a concentration and, linked to this, the lowest average water transparency and the highest water extinction coefficient. It also showed the lowest average oxygen concentration. Finally, the Recovery phase has involved the lowest concentration of nitrate and phosphate and an improvement in the other water quality parameters compared to the Break phase. The activation of the regulation mechanisms seems to manifest through an ammonium production in the water column at the end of the Break phase and low phosphate concentration, probably as a consequence of the activity of ichthyoplankton and the rest of the trophic web.

Negative ecological and socioeconomic effects of eutrophication have been well known for a long time (Vollenweider, 1968; Vollenweider and Kerekes, 1981; United States Environmental Protection Agency [USEPA], 2001; Bricker et al., 2003; Smith, 2003; Jessen et al., 2015), and the concern for their consequences is increasing (Ménesguen and Lacroix, 2018). Moreover, recovery from an eutrophication process is usually difficult and long, and trajectories of impacted ecosystems tend not to be directly reversible, failing to return to the reference status even upon nutrient reduction (Duarte et al., 2008, 2013; Duarte, 2009; Carstensen et al., 2011; McCrackin et al., 2016).

In the Mar Menor, the process of eutrophication has shown three well-defined phases of very different duration that we have been able to observe and characterize (Figure 10).

The start of the process took place in the 1990s due to an increase in the entry of nutrients linked to the change of agricultural practices in the lagoon drainage basin (Pérez-Ruzafa et al., 2002). The comparison between data from 1988, prior to the start of the eutrophication process, and 1997 data, when the process was in its early stages with jellyfish proliferations, showed that the trophic status of the lagoon was changing. Nitrate concentration increased by one order of magnitude, from being always lower than 1 μM NO3− in 1988, to frequently having values higher than 4–6 μM in 1997, particularly during spring and summer (just at the harvest time, when the main irrigations occur in the drainage basin and larger amounts of fertilizers was used in the lagoon’s watershed) (Pérez-Ruzafa et al., 2005a). Furthermore, Pérez-Ruzafa et al. (2005a) found that, while previously to the eutrophication process the nitrate was mainly entering into the lagoon via runoff, as its concentration in the water column showed a positive and significant correlation with accumulated rainfall during the week prior to the sampling day, this positive relationship was lost during the eutrophication process, related to agricultural activity. Although peaks in nutrients inputs occurred, reaching up to 45 μM NO3− (Pérez-Ruzafa et al., 2005a), and still took place after runoff events, this general loss of correlation with rainfall is confirmed in our present work.

On the other hand, P concentrations suffered a sharp decline when the eutrophication process started compared with a decade before. This is explained by the fact that, at the end of 1990, a sanitation plan for urban wastes was established in the lagoon watershed, with the construction of a collector network and water treatment plants. This changed the input regime of this nutrient, moving from direct urban discards with a strongly seasonal pattern, associated with tourism and coast occupation during summer, to an occasional pattern, only when the torrential rains caused the closure of the wastewater treatment plants to avoid their overload (Pérez-Ruzafa et al., 2005a). At the same time, the increase in the nitrate load in the lagoon led to a shift from N- to P-limitation in primary production, maintaining phosphorous concentration permanently low due to its consumption by primary producers.

The highest nitrate concentrations were reached in the springs of 2006 and 2007 with average values exceeding 15 μmol NO3−/L, although the values returned to concentrations under 2 μmol NO3−/L during autumn and winter of the same years. The maximum concentrations of nutrients found in the lagoon water column at that moment were much lower than those measured in the runoff waters that discharge into the lagoon that account for a total input of 219 t N-NO3−/year, 30 t N-NH4+/year, and 52 t/year of total phosphorous (García-Pintado et al., 2007). This difference cannot be explained by an effect of dilution or export to the Mediterranean. The Mar Menor is a choked lagoon with a Water Renewal Time of 318 days (Umgiesser et al., 2014; Ghezzo et al., 2015; Pérez-Ruzafa et al., 2019a). Hydrodynamic models showed that in less than a year the Mar Menor waters would have reached 100% of the concentration contained in the Albujón watercourse waters, which during the Prebreak period represented el 50–60% of total discharges reaching the lagoon (Pérez-Ruzafa, 2010). Therefore, the difference in concentration must be largely attributed to nutrient incorporation into trophic networks through consumption by phytoplankton, macrophytes and microphytobenthos and the biogeochemical cycles inside the lagoon. In this way, surprisingly, despite the high inputs and the nitrate concentration values measured in 2006 and 2007, the mean concentration of chlorophyll a remained low during all the Prebreak phase (1.15 ± 0.02 μg/L) and similar to values recorded before starting the eutrophication process (1.02 ± 0.16 μg/L) (Gilabert, 2001b; Pérez-Ruzafa et al., 2005b; Figure 10).

When the nitrates began to enter in greater amounts in the 1990s, the role played by the jellyfish which proliferated at that time and by the ichthyoplankton resulted in a top-down control of the pelagic system maintaining the chlorophyll a concentration low (Pérez-Ruzafa et al., 2002, 2004, 2005a). This, added to the particular lagoon homeostatic mechanisms showed by the Mar Menor, stopped the main consequences of the eutrophication process for more than 20 years. Such mechanisms are based on (1) a high spatio-temporal hydrographical and biological heterogeneity (Pérez-Ruzafa et al., 2005a, 2007), induced by a restricted connectivity with the Mediterranean sea (Pérez-Ruzafa, 2015; Pérez-Ruzafa et al., 2018, 2019a), (2) a high benthic micro and macrophytes production and filter feeders, scavengers, and detritivores biomass (Pérez-Ruzafa, 1989), and (3) the accumulation of excess production in the sediments or its export out of the system through fisheries and migratory species (Pérez-Ruzafa et al., 2018).

Although some lesser events had already occurred in 2009 and 2010 with chlorophyll peaks exceeding 5 μg/L, the eutrophication crisis and Break phase became evident in 2016 with the sudden increase in chlorophyll a concentration and the generalized loss of water quality in the whole lagoon. During this Break phase, mean chlorophyll a concentration was 7.79 ± 0.22 μg/L, exceeding frequently 10 μg/L.

This translated in a drastic increase of the light extinction coefficient, reaching during the Break phase an average value of 1.24 ± 0.02 /m, that led to the loss of 81% of the macrophyte meadows below 2 m depth compared to the areas estimated in 2008 and 2014 (Pérez-Ruzafa et al., 2012; Belando et al., 2017). This produced a strong social concern and political sensitivity (Limón, 2016) that resulted in the adoption of urgent measures to stop spills of agricultural origin.

According to reported data (Esamur, 2018), from April 2017 to October 2018, the average flow to the Mar Menor through the Albujón water course was reduced to 5.89 ± 1.32 L/s and a load of 38.76 t NO3−/year, representing a reduction of about 97.55% of the flow and 82.3% of nitrate loads regarding the Prebreak and Break periods (García-Pintado et al., 2007; Pérez-Ruzafa, 2010).

At the end of the Recovery phase, mean chlorophyll a concentration was back to levels lower than 2 μg/L (0.94 ± 0.04 μg/L in the last sampled season, Autumn 2018, when light extinction coefficient was 0.42 ± 0.02 /m).

The start of regulation mechanisms in the trophic web, both precluding the chlorophyll a proliferation in 2009 and 2010, and helping the system recover after the 2016 crises, seems to manifest through an ammonium production in the water column, probably as a consequence of the activity of the herbivores and the highest levels of the trophic web. In fact, during the eutrophication process, fishing yields tended to increase in the Mar Menor (Marcos et al., 2015).

The climatic conditions, so far, have not had a major influence on the overall process. Although there are seasonal and inter-annual significant differences in the meteorological and hydrographic parameters, these do not show significant differences between the Prebreak, Break, and Recovery phases of the eutrophication process. Only mean wind velocity in the sampling day had some positive influence on chlorophyll a concentration while wind reaching 15–25 m/s has negative influence. The low or negative relationship between chlorophyll a and nutrients concentration, mainly at small spatio-temporal scales and high nutrient concentrations, is in disagreement with the traditional eutrophication models in which a positive response of chlorophyll a concentration is expected when nutrient concentration increases. This suggests the existence of complex patterns and regulatory mechanisms as proposed by Cloern (2001) or Nixon et al. (2001).

The use of indicators of ecological status in coastal lagoons faces several difficulties. As they are subject to frequent environmental fluctuations and extreme conditions, their functioning in pristine conditions may resemble that of systems under anthropic pressure. Therefore, they are expected to be dominated by opportunistic and r strategist species, giving rise to what has been called the “estuarine quality paradox” (Dauvin, 2007; Elliott and Quintino, 2007; Dauvin and Ruellet, 2009). In any case, this approach for estuaries may not be fully applicable in coastal lagoons (Pérez-Ruzafa et al., 2011b,c, 2013). At the same time, the dynamics of chlorophyll a and nutrients in coastal lagoons can show a high spatio-temporal variability that depends on the observation scale, with a significant space × time interaction and with frequent lags, which complicates the design of sampling programs (Pérez-Ruzafa et al., 2005a, 2007).

Furthermore, these ecosystems show a higher than expected homeostatic capability, showing frequently negative or complex relationships between chlorophyll a concentration and nutrients. So, it has already been established that, contrary to freshwater ecosystems, nutrient concentration is not a good indicator of eutrophication in coastal lagoons, and the same would happen for estuaries and marine coastal areas (Cloern, 2001; Nixon et al., 2001; Boesch, 2002; Bricker et al., 2003; Pérez-Ruzafa et al., 2005a).

This makes it essential to find indicators that detect the critical points of the eutrophication process: (1) in the phases in which the ecosystem is subject to pressures but the functioning of homeostatic mechanisms prevents finding evident deviations in traditional indicators like nutrients or chlorophyll a concentration and (2) in the vicinity of the break points, before the whole system shifts to an alternative steady state and therefore it would be too late to take management measures (Ferreira et al., 2011; Pérez-Ruzafa and Marcos, 2015; Pérez-Ruzafa et al., 2018).

The TRIX index, proposed by Vollenweider et al. (1998), is widely used as an indicator of eutrophication. According to Giovanardi and Vollenweider (2004) and Penna et al. (2004), values ranging from 0 to 4 correspond to high-quality, 4–5 to good, 5–6 to moderate, and 6–10 to degraded conditions. Following these authors, values lower than 4 TRIX units are instead associated with scarcely productive coastal waters, while values lower than 3 are usually found in the open sea. In our work, although this index showed a negative relationship with water transparency (Figure 9B), from 3363 sampling data during the three phases of the eutrophication process, only four samples showed values higher than 4 (ranging between 4.11 and 4.16), therefore indicating high water quality in all the lagoon during the period considered. These highest values were reached in November 2016, just after the starting of the Break phase. This is evidence that, according to the current score scale, it was not able to detect the pressures during the previous 20 years (Salas et al., 2008) nor to anticipate the break point.