Abstract: Perna viridis (P. viridis) has been identified as a good biological indicator in identifying environmental pollution, especially when there are various types of Heavy Metals Accumulations (HMA) inside its tissue. Based on the potential of P. viridis to accumulate heavy metals and the data on its physical properties, this study proffers to determine the relationships between both properties. The similarities of the physical properties are used to mathematical model their relationships, which included the size (length, width, height) and weight (wet and dry) of P. viridis, whilst the heavy metals are focused on concentrations of Pb, Cu, Cr, Cd and Zn. The concentrations of metal elements are detected by using Flame Atomic Adsorption Spectrometry. Results show that the mean concentration of Pb, Cu, Cr, Cd, Zn, length, width, height, wet weight and dry weight are: 1.12±1.00, 2.36±1.65, 2.12±2.74, 0.44±0.41 and 16.52±10.64 mg kg-1 (dry weight), 105.08±14. 35, 41.64±4.64, 28.75±3.92 mm, 14.56±3.30 and 2.37±0.86 g, respectively. It is also found out that the relationships between the Heavy Metals Concentrations (HMA) and the physical properties can be represented using Multiple Linear Regressions (MLR) models, relating that the HMA of Zinc has affected significantly the physical growth properties of P.viridis.
INTRODUCTION
Pollutants as heavy metals can be distributed to aquatic environment through many pathways. These pathways may involve processes, what are also known as bioaccumulation. The accumulation of heavy metals in living tissues, especially mollusc, may cause the increasing of the toxicity levels in living tissues (Dobrowolski and Skowronska, 2001). However, some molluscs, such as P. viridis, have a natural technique to reduce the heavy metals concentration from their tissues which in turn, is related to its dual-shell activity; open and closed feeding method. In addition, P. viridis had also been known as a biological indicator for heavy metals concentration by many researchers (Sivalingam, 1977; Yap et al., 2002; Widmeyer and Bendell-Young, 2007).
According to Widmeyer and Bendell-Young (2007), active feeding behavior of mollusc may increase the concentrations availability of pollutant in its tissue. Mollusc is also exposed to different food suspensions consisting mixtures of sediment, particulate matter and seston. The different concentrations of pollutant in aquatic environment may affect the growth of mollusc in aquatic environment. So, the purpose of this study is to determine the relationships between the heavy metals accumulations in total soft tissue with the physical properties of P. viridis.
MATERIALS AND METHODS
Study site: Mengkabong Lagoon is located in the Tuaran District, which is 53 km away from the city of Kota Kinabalu in Sabah. It is dominated by a mangrove ecosystem which is suitable for aquaculture activities (Fig. 1). There are four stations of aquaculture activities being identified, namely B1, B2, B3 and B4, as shown in Fig. 1.
Data samplings of P. viridis: Fresh samples (n = 120) of P. viridis in Fig. 2 are collected randomly in different length sizes, ranging from 60 mm to 113 mm. Each individual is separated from its shell and the constant weights of tissue samples are taken at 60°C (Silva et al., 2006; Blackmore, 2001; Nair et al., 1993).
The digestion method of tissue samples is carried out by first adding 10 mL of concentrated nitric acid to it and then heated on a hotplate at 70°C (Silva et al., 2006; Yap et al., 2002; Blackmore, 2001).
| Fig. 1(a-b): | Study sites at Mengkabong Lagoon in Tuaran district, Sabah (Scale 1 cm: 500 m) |
| Fig. 2: | Schematic size measurement of P. viridis |
Distilled water is then added to it up to 50 mL. The heavy metals contents are then measured using FAAS after filtering it with a 45 μm sized Withman membrane. The SI unit for the concentration of heavy metals in the tissues is presented in mg kg-1 (dry weight).
| Table 1: | Concentration of heavy metals in a standard reference material, lobster hepatopancreas (TORT-2) |
| All data as Mean±SE in mg kg-1 dry weight (n = 5) | |
The accuracy and precision of the procedures are compared to a standard reference material, Lobster Hepatopancreas (TORT-2) which is provided by the National Research Council of Canada. The recovery ranges from 89 to 108%, as shown in Table 1.
| Fig. 3: | Procedures in mathematical modeling phases using Multiple Linear Regressions (MLR) |
Normality, multicollinearity and model-building: The variables are initially tested for their normality distributions and are based on the Kolmogorov-Smirnov statistics with Lilliefors significance level of more than 0.05, since the sample size, n = 120 is large (>50). For non-normal data, appropriate transformations has to be done first, followed by minimizing collinearity between the variables (if there exist), based on the Pearson Correlation Coefficient matrix. Next, follows the model-building procedures which involve the following phases, as shown in Fig. 3, before finally the selection of the model equations representing the relationships between the heavy metals concentration with the respective physical properties.
RESULTS
Appropriate transformations have been carried out to normalize the data variables. Table 2 shows the definitions of the newly transformed variables and its units used. New variables are assigned to it for model simplicity.
Table 3 depicts the normality tests of the variables before and after transformation, showing that most of the variables have turned to normal after transformation. However, only three of the variables are not normal, although they have undergone their best transformations. Their significant p-values do not exceed 0.05 to indicate normality, but their normality plots exhibit nearly normal distributions.
Table 4 depicts the Pearson Correlation Coefficient matrix of the transformed variables model say M31. Any existence of multicollinearity between the variables has to be remedied first before further analysis can be done. It can be seen that there are no absolute coefficient values of |r|≥0.95 exist in the model. Therefore, no multicollinearity effects exist in this model function.
The number of the all possible models would be given by:
where, ‘q’ would be the number of single independent variables. Table 5 depicts some selected models from the thirty-one (31) possible models for Lead (Pb) for the five single independent variables. The all possible models for Lead are depicted in the Appendix. Simultaneously, similar models are also applied to the other heavy metals concentrations. Hence, overall there are 155 models to be selected from in relating the HMA as the dependent and the physical growth factors as the independent variables. However, in this work the models with interactions between the independent variables are not carried out.
Table 6 below depicts the Pearson Correlation Coefficient matrix for model M31 of Pb. No multicollinearity exists, so the procedures of removing insignificant variables will then be carried out using backward elimination method of the Coefficient test. The number of eliminated insignificant variables will then follow after the parent model as shown by Abdullah et al. (2008).
Table 7 tabulates the results of the best mathematical models which represent the relationships of the heavy metals accumulation versus the physical properties of P. viridis. Comparisons of the best models after elimination amongst the heavy metals would show that Pb has a negative relationship with the wet weight, while Cu and Cd have negative relationships with the dry weight properties. In addition, Zn has positive relationships with two of the physical properties, i.e., the width and dry weight, while Cr has a good relationship with three of them, namely, the width, height and dry weight properties. Further comparisons based on the least sum of square Errors (SSE), shows that Zn, is a better indicator amongst all the heavy metals, follow by Pb and Cr.
| Table 2: | Transformations of Defined Variables (heavy metals concentration and physical properties) |
| Table 3: | Normality tests before and after transformations of heavy metals concentrations and physical properties |
| Table 4: | Pearson correlation coefficient matrix of transformed variables (heavy metals concentration and physical properties of P. viridis) |
| Table 5: | Some selected possible models for lead (Pb) |
The model equation is transformed back into its defined variables as indicated and is given by:
In other words, four HMA dependent variables have shown a good relationship to the dry weight physical independent properties, except for Pb which is negatively related to the wet weight physical properties.
| Table 6: | Coefficient matrix of model M31 for Pb |
It can also be seen from the models that the length of P. viridis, has not become one of the determining indicator of the physical properties for HMA identification. Thus, it can be said that the HMA has been affected significantly by the physical growth properties, especially the width, height, wet and dry weight properties of P. viridis.
After deriving the mathematical model which can represent the best indicator for HMA, the goodness-of-fit tests (i.e., the normality test and randomness test) are carried out on the standardized residuals so as to ascertain the validity of the derived model. The choice of the best mathematical model is based on the least sum of square error (SSE) where the assumptions of normality and linearity of the model residuals are verified. Figure 4 above indicates that these assumptions of the goodness-of-fit the best mathematical model represented by Zn have been met.
| Table 7: | Mathematical models equations for Y1-Pb, Y1-Cu. Y3-Cr, Y4-Cd, Y5-Zn |
| Fig. 4(a-b): | Residuals histogram plot and residuals scatter plot for Zn (a) Histogram and (b) Scatter plot, Dependent variable: z-0.22 |
DISCUSSION
The experimental results show that P. viridis has accumulated the Heavy Metals Concentration (HMA) in the following order: Zn>Cu>Cr>Pb>Cd. As a result, high accumulations of Zn and Cu have indicated the high potentiality of these elements being accumulated in the tissue bodies of P. viridis. According to Dobrowolski and Skowronska (2001), mollusc was found to be a very effective organism to accumulate heavy metals compared to fish. In addition, Blackmore (2001) had also reported that high accumulation of Zn and Cu were also found in Saccostrea cucullata.
The mathematical models of the respective heavy metals (Pb, Cu, Cr, Cd, Zn) in Table 7, show that different heavy metal would affect differently the physical properties. After the procedures of all the phases in model-building have been carried out, the best model equation shows that Zn is the best indicator based on its least SSE value, compared to the other metals. The model also indicates that Zn has a directly positive relationship with the variable, dry weight and inversely proportional to the width of P. viridis. The best model with the highest SSE from Table 7, represented by Cu, however, has also shown that dry weight too is a determining factor. These results are also in accordance with Boyden (1974) and Otchere (2003), where dry weight and sizes of each individual mollusc, are important components when measuring the concentration of heavy metals accumulated in its body.
CONCLUSION
P. viridis has accumulated high concentration of Zn and Cu but it is still in the permissible safety levels for human consumption, especially as a seafood resource (100 and 30 mg kg-1). There are significant relationships found between the Heavy Metals Concentrations (HMA) and the physical properties. These relationships can be represented using Multiple Linear Regressions (MLR) models. The models thus relate that the HMA has affected significantly the physical growth properties, especially by the width, height, wet and dry weight properties of P. viridis. Further works can hence be done by looking at the interactions between the independent variables. Comparisons of the best model equations based on the least SSE, indicate that Zn is the best indicator and is given by the model equation:
ACKNOWLEDGMENT
The authors would like to acknowledge the financial support from Universiti Malaysia Sabah in this study.
| Appendix: | All possible models with single independent variables |