The nutrient concentrations in the water column were processed following the protocols by Grasshoff et al. (1983) and Parsons et al. (1984) and measured by means of UV/VIS Spectrophotometer (U-1800, Hitachi). The percentage of labile OM concentration was estimated using the protocols by Loh et al. (2008). The concentrations of chlorophyll-a and POC were assessed in both the water column and the sediments following the methods of Hedges and Stern (1984) and Yentsch and Menzel (1963)measured by means of Fluorometer (TD-700, Turner Design). The granulometry samples were analyzed following the protocols by Gray and Elliott (2009). The samples of TRIS were analyzed following the hot distillation method (Kallmeyer et al. 2004). The final measurement of the TRIS values was based on the method reported by Cline (1969).

The heavy metal samples were treated following the methods described by Kalantzi et al. (2013). Data quality assurance was performed by using one blank and one certified reference material (marine sediment, NCS DC75305 and NCS DC75301) from the China National Analysis Centre in every 6 samples digested. For the samples from the sampling cruise of February 2011, the average recoveries of all elements of NCS DC75305 was 92.61±1.43% (n=33) and of NCS DC75301 was 85.67±4.64% (n=31). For precision assessment, three different sediment samples were analyzed 3 times each and RSD was lower than 12% for all elements except for Cd, which had RSD ~13.5%. The element concentrations in the digestion blanks were typically very low and were subtracted from the sample values. The limits of detection (LOD) of the procedure were calculated by multiplying the standard deviation of the blanks (n=6) by three and were: 0.11 (Li), 0.02 (Be), 15.49 (Na), 12.25 (Mg), 0 (Al), 17.94 (P), 31.46 (K), 16.43 (Ca), 0.04 (Sc), 0.02 (V), 0.3 (Cr), 0.16 (Mn), 24.49 (Fe), 0.03 (Co), 0.35 (Ni), 0.5 (Cu), 0.76 (Zn), 0 (Ga), 0.02 (Ge), 0.14 (As), 0.01 (Rb), 0.06 (Sr), 0.01 (Y), 0 (Pd), 0.02 (Ag), 0.01 (Cd), 0 (Cs), 0.01 (La), 0.01 (Ce), 0 (Pr), 0 (Nd), 0 (Eu), 0 (Sm), 0 (Gb), 0 (Tb), 0 (Dy), 0 (Ho), 0.01 (Er), 0 (Tm), 0 (Yb), 0 (Lu), 0 (Tl), 0.03 (Pb), 0 (Th) and 0 (U) mg/kg.

The rest of the sediment samples were processed in a different analytical run. Average recoveries of all elements of NCS DC75305 was 96.7±2% (n=41) and of NCS DC75301 was 91.6±4.7% (n=39). For precision assessment, three different sediment samples were analyzed 3 times each and RSD was lower than 12% for all elements. The element concentrations in the digestion blanks were typically very low and were subtracted from the sample values. The limits of detection (LOD) of the procedure were calculated by multiplying the standard deviation of the blanks (n=14) by three and were: 4.67 (Li), 0.01 (Be), 35.28 (Na), 15.37 (Mg), 10.76 (Al), 22.78 (P), 68.73 (K), 132.70 (Ca), 0.04 (Sc), 0.01 (V), 0.3 (Cr), 0.46 (Mn), 34.09 (Fe), 0.01 (Co), 2.28 (Ni), 0.38 (Cu), 3.82 (Zn), 0 (Ga), 0.02 (Ge), 0.06 (As), 0.06 (Rb), 0.36 (Sr), 0.01 (Y), 0.01 (Pd), 0.01 (Ag), 0 (Cd), 0 (Cs), 0.03 (La), 0.05 (Ce), 0.01 (Pr), 0.02 (Nd), 0 (Eu), 0.01 (Sm), 0 (Gb), 0 (Tb), 0 (Dy), 0 (Ho), 0 (Er), 0 (Tm), 0 (Yb), 0 (Lu), 0 (Tl), 0.09 (Pb), 0.01 (Th) and 0 (U) mg/kg.

Both the analysis of TRIS and element concentrations were realized in the in the Marine Ecology Laboratory of the University of Crete.

A series of methods were used for the statistical analysis of the data. Matrices of the abiotic parameters over stations and seasons were produced to compare the similarities: i) between all stations over all seasons, ii) between the stations for each season, iii) between the stations over seasons per lagoon, iv) between the stations near to the sea and the stations in the innermost parts of the lagoons. The similarities were calculated using the Euclidean distance and they were used to create nMDS plots. Permutational multivariate analysis of variance (PERMANOVA) (Anderson 2001) was performed to test for the significance of the factors affecting the resulting similarity patterns. A reduced model was used for the permutation of residuals combined with a Type III (partial) sum of squares.

The Principal Component Analysis (PCA) was used to detect the parameters responsible for the patterns observed between the stations. The similarities between the stations were estimated again using the Euclidean distances. For the previously described routines the PRIMER v. 6.1.8 (Clarke and Gorley 2006) software was used. Significance of the PCA analysis was tested with the Kaiser-Guttman approach (Jackson 1993). The approach calculates the ratio λ between each eigenvalue and the average of all eigenvalues. The information from the components with λ>1 is the one retained. The value to consider a variable as strongly associated with any of the eigenvectors was set at higher than 0.5 or lower than -0.5.

Moreover, linear regression was applied between the values of salinity (independent variable) and those of the abiotic parameters (dependent variables) (Xu et al. 2008), to identify the most important parameters per lagoon affected by the seasonal inflows or evaporation. The Kolmogorov-Smirnov normality test was applied on the abiotic parameters and for those non-fitting the normal distribution, the logarithmic transformation was used prior to enter analysis.

The heavy metal concentrations in the sediments were measured per station and the outcomes were compared against the Threshold Effect Level (TEL) using the standards implemented by the Canadian Freshwater Sediment Guidelines (Burton 2002) and against the Sediment Quality Guidelines (SQG) implemented by the United States Environmental Protection Agency (USEPA) (Pekey et al. 2004). Moreover, the geo-accumulation index (I_geo) (Muller 1969) was implemented in order to estimate the heavy metal concentrations above background concentrations. The index values are ascribed to seven enrichment classes. Each class is attributed to an estimate of the heavy metal pollution level (Muller 1969). For the geochemical background values, the reference of Martin and Meybeck (1979) was used. Additionally, the contamination factor (C_f) for seven pollutants (As, Cd, Cu, Cr, Pb, Zn as suggested by Hakanson (1980) and Ni that showed high sediment concentrations) was calculated, along with the degree of contamination for each station (C_d). For the contamination factor, the same background concentrations as for the I_geo index were used. The values of the C_f factor and the C_d degree correspond to different intensity of contamination (Hakanson 1980).

The annual fluctuations of the abiotic variables per station and season are given in the Supplementary material (Suppl. material 1)

The nMDS ordination analysis, applied on the values per station over all seasons, revealed the multivariate similarity patterns based on the values of the abiotic variables (Fig. 2). The PERMANOVA analysis supports clustering of the stations by season (pseudo-F=9.048, p=0.001), by lagoon (pseudo-F=4.018, p=0.001) and by lagoon over season (pseudo-F=2.256, p=0.001). The PCA performed on the same dataset showed that 51.1 % of the total variation was described by the first three PCA axes (λ>1) but no distinguishable eigenvector was observed for any of the abiotic variables tested (Fig. 3).

Figure 2.  

nMDS analysis plot between the stations based on the abiotic data from all the lagoons and sampling seasons.

Figure 3.  

PCA plot of the stations based on the environmental data from all the lagoons and sampling seasons.

The results of the nMDS analysis applied on the data per station over single seasons showed no significant grouping between the stations in autumn and spring, while in winter and summer (Fig. 4) there was clear assembly of the stations from the same lagoon (PERMANOVA: pseudo-F=2.949, p<0.05 for winter; pseudo-F=2.384, p<0.05 for summer). The distance from the sea was not found to play any role on the sample clustering.

a

b

c

d

Figure 4.

nMDS analysis plots between the stations based on the abiotic data during:

a: Autumn  
b: Winter  
c: Spring  
d: Summer  

Applied on the same datasets, the PCA analysis revealed the first two PCA axes (λ>1) identified accounting for the 63.2% of the total variation in the autumn samples. The first principal component was not found to be strongly associated with any of the variables. On the contrary, the second axis was associated (0.509) to the CPE of the sediment (Fig. 5a). For the winter samples the 73% of the variability was explained by the first two axes (λ>1) (Fig. 5b). The first principal component was strongly correlated to the chl-a of the water column (-0.50), while the second component was strongly associated with the NO₃ (-0.507). The 65% of the variability was expressed by the first two components (λ>1) in PCA from the spring stations (Fig. 5c). Nevertheless, only the first one was found to be strongly associated to the following variables: the percentage of silt and clay (0.517) and the percentage of sand in the sediment (-0.517). In the PCA with the data from the summer sampling, only the first two axes were found to fulfill the Kaiser-Guttman criterion and they described the 64.8% of the total variability. No variable was associated to the first principal component, however, the second axis was characterized by the NH₄ (-0.677) (Fig. 5d).

a

b

c

d

Figure 5.

PCA plots of the stations based on the environmental data of all the stations during:

a: Autumn  
b: Winter  
c: Spring  
d: Summer  

The nMDS analysis was also applied on data deriving from each lagoon, separately and over all seasons (Fig. 6). The results did show significant grouping of the stations except for Rodia lagoon. In all other lagoons PERMANOVA supported the grouping of stations following a seasonal pattern: pseudo-F=4.952, p<0.05 for Mazoma lagoon; pseudo-F=4.802, p<0.05 for Logarou lagoon; pseudo-F=3.647, p<0.05 for Tsopeli lagoon; pseudo-F=2.839, p<0.05 for Tsoukalio lagoon. The stations did not appear to be grouped according to their distance from the sea in any of the cases described above.

a

b

c

d

e

Figure 6.

nMDS analysis plots between the stations based on the abiotic data of four sampling seasons from:

a: Mazoma  
b: Logarou  
c: Tsopeli  
d: Tsoukalio  
e: Rodia  

The results of PCA analysis applied on the data from Mazoma lagoon (Fig. 7a), showed that the 70.3% of the total variability was explained by the first two axes, which met the Kaiser-Guttman criterion. However, no variable was strongly associated with any of the components. The same analysis for Logarou lagoon indicated the first three PCA components (λ>1) to represent the 86.7% of the variability (Fig. 7b). Similarly to Mazoma, no variable was strongly associated to any of the three axes. For the data from Tsopeli lagoon, the highest percentage of the total variation was explained by the first two principal components (65.6%). The first component was strongly associated with the phaeophytin in the water column (-0.51), the second one by NO₂ (-0.524) (Fig. 7c). Similarly, for the data from Tsoukalio lagoon the first two axes (λ>1) accounted for the 67.2% of the variability (Fig. 7d). The NO₃ was strongly associated (-0.516) to the first component, the second component was strongly related to the ratio of chl-a/phaeophytin in the water (-0.683). Finally, for the samples from Rodia, the results of the PCA showed the first two components (λ>1) to represent the 71.3% of the total variability, but no variable was found to be strongly associated (Fig. 7e).

a

b

c

d

e

Figure 7.

PCA plots of the stations based on the environmental data of four sampling seasons from:

a: Mazoma  
b: Logarou  
c: Tsopeli  
d: Tsoukalio  
e: Rodia  

Again, the distance of the stations from the sea, did not appear to be related with any of the vectors identified by the PCA.

All the abiotic variables were plotted against salinity in each lagoon, to give an estimate of the concentration ranges in relation to the freshwater inflows. The linear regression plots of the parameters significantly related to the salinity shifts, are provided in Suppl. material 1. In Mazoma lagoon, the concentration of NH₄ in the water (R²=0.53; p<0.05), phaeophytin (R²=0.63; p<0.05) and CPE (R²=0.55; p<0.05) in the sediment showed significant increase, by increasing salinity, while the concentration of O₂ (R²=0.77; p<0.05) in the water column appeared to be decreasing with increasing salinity. The same test for Logarou lagoon, indicated significant reduction (R²=0.91; p<0.05) of the water POC concentration values as the salinity increased. The picture was different for Tsopeli lagoon, where three environmental factors presented significant changes by the increasing salinity values: temperature in the water column (R²=0.73; p<0.05) was augmenting, while the concentration of NO₂ (R²=0.52; p<0.05) and O₂ (R²=0.94; p<0.05) in the water were significantly lowered with increasing salinity. In Tsoukalio lagoon, the concentrations of SiO₂ (R²=0.52; p<0.05) in the water and chl-a in the sediments (R²=0.50; p<0.05) were higher as salinity was increasing. On the contrary, the values of O₂ (R²=0.70; p<0.05) in the water along with the values of Eh (R²=0.52; p<0.05) in the sediments were reducing with the rising of salinity. Finally, the concentration decrement of O₂ (R²=0.85; p<0.05) in the water was the factor that showed significant shifts as the salinity was increasing, in Rodia lagoon.

The concentrations of seven of the heavy metals analyzed were compared between the sampling stations (Figs 8, 9). The As concentrations were found to exceed the SQG threshold (3 ppm) in all the stations except from the outer station of Logarou (2.38 ppm) but they never transcended the TEL limit (5.9 ppm). Similarly, the concentration of Cu was found to be above the SQG level (25 ppm) and, in the cases of the inner stations of Mazoma (45.87 ppm) and Logarou (36.51 ppm), even higher than the TEL level (35.7 ppm). The lowest values of Cu were recorded to the outer station of Rodia lagoon (14.48 ppm). The results were different for Cr and Ni. The concentrations of these elements were found to be high in all the stations and exceeding the levels of both SQG (SQG_Cr=25; SQG_Ni=20) and TEL (TEL_Cr=37.3; TEL_Ni=18).

a

b

c

d

e

f

Figure 8.

Heavy metals concentrations in the sediment from each station. The dashed line represents the TEL threshold. The continuous line represents the SQG threshold. Abbreviations correspond to: M: the stations of Mazoma lagoon, LO: the stations from Logarou lagoon, TSO: the stations from Tsopeli lagoon, T: the stations from Tsoukalio, R: the station from Rodia lagoon, Channel: the outer station of Rodia lagoon.

Figure 9.  

Zn concentrations in the sediment from each station. The dashed line represents the TEL threshold. The continuous line represents the SQG threshold. Abbreviations of the stations as in Fig. 8.

The results of the I_geo index classified most of the stations to the zero class of the uncontaminated sediments (Table 2). The inner station of Mazoma was the exception in this pattern, as it was ranked to class one which includes the level of of the uncontaminated to moderately contaminated sediments, based on the Cd and Pb and to class two, which includes the moderately contaminated sediments based on the Cr and Ni. However, the index did not rank sediments of this station as contaminated by As, Cu and Zn. The sediments were classified to low or moderate Cr contamination for all the stations and to low and moderate Ni contamination for most of the stations. Overall, the index was high for Cr and Ni and low for Zn and Pb. The order of the elements as ranked by the I_geo index values was Cr>Ni>Cd>Cu>As>Zn>Pb.

Similar results were derived from the calculation of the contamination factor (C_f) (Fig. 10). The factor for the inner station of Mazoma was found to be above the level of low contamination for the Cd, Cu, and Pb and above the moderate contamination level for Cr and Ni. The values were exceeding the low contamination threshold for Cd in both of the Tsopeli stations and for Cu and Pb in the sediment of the inner station from Logarou lagoon. The values of Cr and Ni were over the low level of contamination in all the sediment samples.

a

b

c

Figure 10.

The C_f index values calculated for each element per station. Abbreviations for stations are the same as described in Fig. 8. The dashed, continuous and dotted lines represent the thresholds for low, moderate and high contamination of the sediment, respectively

a: index values for the As, Cd, and Cu  
b: index values for Pb and Zn  
c: index values for Cr and Ni  

The contamination degree was calculated for all seven metals (Fig. 11a) and the mean contamination degree was estimated for all the stations (Fig. 11b). Results of the index presented only three stations under the level of low contamination: the outer stations of Mazoma and Logarou and the inner station of Rodia. According to the average values of the C_d (mC_d) the stations that were ranked as with the highest contamination for all the contaminants were the inner stations of Mazoma and Logarou. The elements with the highest C_d values were Cr an Ni in the sediments over all stations. The order of the contaminants following the degree of contamination in the sediments were Cr>Ni>Cd>Cu>As>Zn>Pb.

a

b

Figure 11.

The dashed and continuous lines represent the thresholds for low and moderate contamination of the sediment, respectively.

a: The Contamination degree values for each element  
b: The mean Contamination degree values per station  

Overall, no evidence of intense disturbance was observed in patterns derived by the abiotic variables in the lagoons under study. Salinity ranges were negligible between the inner and the outer parts of the lagoons in all seasons. The annual pattern, however, was characterized by high salinity values in the summer and autumn and low in winter and spring. Salinity values follow a pattern similar to those described in the same lagoons from older studies (e.g. Kormas et al. 2001, Katselis et al. 2013), but also in other lagoons from Greece (Kevrekidis 2004, Orfanidis et al. 2005) and the Mediterranean Sea (De Casabianca et al. 1997, Vignes et al. 2009). However, salinity levels in Tsoukalio and Rodia were found much lower than all the other lagoons (5.6-20 psu in Tsoukalio; 4.7-23.4 psu in Rodia). Kormas et al. (2001) have reported higher salinity levels, ranging between 13-28.9 psu in Tsoukalio and 11.6-25 psu in Rodia. Moreover, Nicοlaidou et al. (2005) mentioned salinity ranges between 14.0-36.5 psu for Tsoukalio and 5.0-35.0 psu for Rodia. The low levels of salinity measured in the two lagoons during the study are comparable to the ranges referenced from the Evros Delta (4.0-25.0 psu; Nicοlaidou et al. 2005), and are considered as indicative of increased freshwater inflows. Furthermore, the salinity levels in Tsopeli (14.6-42.1 psu) were found to have a wider range than the ones remarked in the study of Nicοlaidou et al. (2005) (21-38 psu) and the paper of Reizopoulou et al. (1996) (21-35 psu).

The oxygen concentration shifts during the year, were normal for all lagoonal systems studied. Lower concentrations were measured during summer (1.4 - 6.32 mg/l) and autumn (3.66 - 6.45 mg/l), whereas higher values were observed in winter (7 - 9.43 mg/l) and spring (6.85 - 11.7 mg/l). A similar pattern has been observed in Monolimni lagoon (Kevrekidis 2004). Older studies in the lagoons of Amvrakikos reveal high values during autumn (10.4 - 10.7 mg/l) and lower ones (6 - 9.2 mg/l) during the rest of the year (Kormas et al. 2001), but the ranges were generally in accordance with the measurements reported from previous studies in the lagoons (2.8 - 12.1 mg/l) (Reizopoulou and Nicolaidou 2004). The oxygen concentration fluctuations are influenced by the salinity and temperature ranges and by the biological activity (Best et al. 2007). The annual DO pattern observed in the lagoons of Amvrakikos could be explained by the salinity and temperature annual profiles, as increased levels of temperature and salinity inhibit solubility of the dissolved oxygen (Best et al. 2007).

The nMDS plots and PERMANOVA analysis for all stations and over all sampling seasons revealed significant differentiation between the lagoons, between the seasons and also between the lagoons and over the seasons. However, the location of the stations within each lagoon does not appear to be an important factor. This could be attributed to the physicochemical characteristics of each lagoon, as the PCA analysis showed no variable playing a predominant role in the pattern of variation among stations. This pattern seems to change when the data are treated per season. Although no specific grouping of stations was observed during autumn, the concentration of CPE in the sediments seemed to influence their arrangement along the PCA axes. Overall, the CPE levels in the sediment were high. However, in the case of Mazoma, the stations were found to have higher concentrations, which were increasing with the increasing salinity. The percentage of chl-a in autumn was about 40%, suggesting that although there was continuous input of primary organic matter in the sediment, it was mostly consisted of accumulated chl-a (Pusceddu et al. 1999, Mirto et al. 2004, Sestanovic et al. 2009).

In Mazoma lagoon, the concentration of NH₄ was also found to increase with the increasing salinity, being highest during summer. This response is anticipated when the degradation of the organic matter in the sediment takes place (Bonanni et al. 1992, Diamantopoulou et al. 2008). However, the summer levels were extremely high in respect to the ones reported from other Mediterranean lagoons (Orfanidis et al. 2005, Cobelo-García et al. 2012, Guerra et al. 2013). The degradation of organic matter in the summer, is also supported by the significant decrease of the oxygen concentration in the water column with the increasing salinity levels. Increased NH₄ levels are also noticed during hypoxic events (Wu 2002), however the oxygen concentration (4.76 - 9.43 mg/l) was measured to be within the natural ranges (4.8-6.8 mg/l) (Nicοlaidou et al. 2005).

Only the POC concentration in the water column was found to be significantly correlated with the salinity change in Logarou lagoon. The highest concentrations were noticed in winter and they seem to follow the elevated levels of chl-a in the water column, over the same period, suggesting phytoplanktonic origin of POC.

In Tsopeli lagoon, the PCA analysis pointed out the concentration of phaeophytin and NO₂ in the water as the most important factors for the stations ordination. The NO₂ concentrations were also found to be increasing with the declining salinity levels, indicating inflows through the freshwater. The oxygen concentration annual shifts were following the same pattern as the NO₂. Oxygen concentration presented the lowest values in this lagoon during summer and autumn, however this seems to be natural for Tsopeli as such a decline has been observed before (Reizopoulou and Nicolaidou 2004).

For Tsoukalio lagoon, the PCA analysis showed that the NO₃ concentration and the ratio chl-a/phaeophytin in the water column were playing an important role on the station pattern throughout the year. The chl-a/phaeophytin ratio takes its highest values in autumn. Although the CPE in the water was not high, the percentage of chl-a was assessed to reach over 90% of the CPE against the phaeopigment content in autumn, dramatically dropped under 50% in winter and even lower than 30% in spring (to the outer stations). Such a decline of chl-a concentrations in the water column, along with a concurrent increase of phaopigments, which are considered as chl-a degradation products, is often attributed to intense zooplankton grazing activity (Taguchi et al. 1993).

Similarly to Logarou, fluctuations of the abiotic factors in Rodia seemed to have lower ranges. Oxygen concentration was the only factor found to change significantly in the lagoon, with the increasing salinity levels. The lowest values were measured in the lagoon during summer, nevertheless they were not found to be out of the ranges described for the lagoon by previous studies (Kormas et al. 2001).

The concentrations of heavy metals were higher than the SQG threshold in many cases, but lower than the TEL threshold for the majority of the elements. However, the concentrations of Cr and Ni were exceeding both of the limits in all the stations. The same elements were classified to the class one of the index for all the lagoons and they were surpassing the level of low sediment contamination of the C_d index. The mean contamination factor was exceeding the lower limit of contamination for all the stations, but none of them was over the threshold of moderate contamination. However, the concentrations of these elements in the lagoonal sediments were found in lower levels than those reported by Christophoridis et al. 2007 in the same lagoons. Karageorgis (2007) reported also high values of Cr and Ni concentration in the sediments of Tsopeli, Rodia, Tsoukalio and Logarou lagoons. The author argued that the elements were contained in the soil minerals of Arachthos River and they are transported in the lagoons through the river inflows.

No evidence of severe disturbance was detected in the lagoons of Amvrakikos. The annual fluctuations of the parameters were following the profiles reported before in the area, suggesting that the physicochemical functions in the lagoons do not suffer any major impact. The annual oxygen fluctuations were found to be significant in most of the lagoons studied. Oxygen concentration values were higher during winter and spring and were severely dropping in summer and autumn, following the pattern already described by other authors for the Mediterranean lagoons (e.g. Sfriso et al. 1992, De Casabianca et al. 1997). However, there were no indications of dystrophic crises events in the samples collected from the Amvrakikos Gulf lagoons and the analysis of the heavy metal concentrations showed no evidence of severe contamination. The lagoons are exploited for their high fish production. Therefore, their optimal environmental conditions are important not only for the welfare of the ecosystem, but also for the prosperity of the local community.