The application of chemometrics in metals source of identification in Brunei Bay surface sediment.

Brunei Bay is a unique ecosystem which offers a vast biodiversity. This study was carried out to define the source of metals in the surface sediment of Brunei Bay to ensure the bay’s health. The secondary data were analysed using chemometrics analysis to verify the possible factors that influence metals distribution in Brunei Bay sediment. Samples were collected several times during 2013 to 2014 using Ponar grab at 16 stations within the bay. Samples were then dried, pre-treated, digested and analysed using Inductively Coupled Plasma Mass Spectrometry (ICPMS) in the laboratory. Overall, the mean concentration of metal, sediment pH and clay fraction were significantly changed during different sampling periods, as the changes were presumed affected by seasonal changes. The Pearson correlation has pointed that metals were dominantly derived by natural input; however, the total organic carbon was proven to be derived by anthropogenic sources. Moreover, the principal component analysis has verified that the distribution of metals in the bay’s sediment was dominantly influenced by natural processes. However, the utilization and manipulation of marine resources are slightly affecting the bay’s ecosystem which may deteriorate the ecosystem health soon.


Introduction
Metals are naturally-occurring elements in the Earth's crust, derived by natural processes such as rock weathering, eruption of volcano and forest fires (Masindi & Muedi, 2018). Along with the rapid urbanization, anthropogenic metals are degrading the environmental health (Zhang et al., 2012;Fashola et al., 2016;Gabarrón et al., 2017). Anthropogenic metals have developed serious problems due to their toxicity, non-degradable and bioaccumulation behaviour (Xiao et al., 2019a;Xiao et al., 2019b). They can affect the survival, diversity and richness of aquatic species, which will affect the environment biodiversity (Achary et al., 2016).
Some metals are essential elements for living organisms and are needed for biochemical function and physiological mechanism. However, metals can be noxious when their threshold concentration exceeded. On the other hand, some trace metals are toxic to living organisms and deleterious to environmental health even at a low concentration. Excessive content of metals in the water column may cause metals magnify to a higher trophic level in the food chain through dietary, later bio-accumulate and will led to various health effects (Wei et al., 2014;Gabarrón et al., 2017). For example, high exposure of mercury and cadmium has caused severe respiratory and neurological disorders in human and animals such as Minamata and Itai-itai disease (Shagirtha et al., 2011;Bernhoft, 2013;Jaishankar et al., 2014).
Metals are highly persistent and non-biodegradable. Regardless of bio-accumulating in living organisms, excessive metals also can be settled down onto the sediment bed (Malferrari et al., 2009). Since metals are non-biodegradable compounds; therefore, they permanently remain in the sediment bed for a long period of time (Kumar et al., 2013;Zahra et al., 2014;Zeng et al., 2019). The past research has discovered that metals pollution in the sediment caused the loss of sediment quality Rabaoui et al., 2015). Moreover, such event will also result in the presence of red tides, degradation of some marine species, unstable of different local benthic communities and malformations in different benthic foraminifers' species (Rabaoui et al., 2015;Ayadi et al., 2016;El Kateb et al., 2016).
Sediment is pollutants sink which can preserve environmental changes history record since the past billion years (Chuan & Yunus, 2019). Numerous researches in the worldwide have been using sediment to study the impact of urbanization and anthropogenic activities towards the aquatic environment since the pollutants are discharged into the water column and will be accumulated in the sediments (Malferrari et al., 2009). Through various processes of remobilization under variable conditions, sediment bound metals may be released into water body again, thus intensifying the potential of heavy metals entering food chain and food web (Ong et al., 2015). Since metals are toxic, non-biodegradable compounds can accumulate in sediments and living organisms; therefore, determine the source of metals is important and need to be investigated utterly, to reduce the degradation of environment for future needs.
Chemometrics is known as the application of multivariate statistical modelling on chemistry data (Simeonov et al., 2002). Through different environmental technique application, chemometrics has become an important tool for classification of samples identification of pollution sources. Throughout this application, complex relationships in environmental science area can be reveal; plus can reduce the cost of sample collection and analyse (Şehribanoğlu et al., 2020). Several analyses that are commonly employed to extract maximum relevant chemical information are the principal component analysis (PCA) and discriminant analysis (DA) (Silva Junior et al., 2020). The principal component analysis is unsupervised pattern recognition technique that explains inter-relationship between variables by reducing the dimensionality of large datasets. It is actively used in environmental science research due to its efficiency in reducing error and data loss in the same time (Jolliffe and Cadima, 2016). This analysis is quite similar with correlation or regression analysis method, as it defines correlation of the variable's principal components (PCs) by cutting off the data dimensional (Wunderlin, 2001;Simeonov, 2002;Liu et al., 2003). PCA is widely applied in different areas by researcher due to its considerable reduction in the number of variables and the detection of structure in the relationships of different variables. In the meantime, the discriminant analysis provides statistical classification of samples based on their common attributes. It builds up discriminant function for each group by computing the raw data, which will then determine the variables that discriminate between variable groups (Singh et al., 2005). The present study was carried out to further investigate the source of metals in the Brunei Bay's surface sediment for a better view on the influence of anthropogenic activities towards the ecosystem, by analysing the variable's correlation using PCA and determining the discriminated variables using DA.

Description of the sampling location
Brunei Bay is located at the northwest region of the Borneo Island and encompass by Labuan Island, Sabah, Sarawak and Brunei. This bay has an extensive seagrass pasture and thus acting as important foraging ground for green turtles in Southeast Asia . Additionally, according to Jaaman et al., (2008), Brunei Bay serves as habitats for several species of cetaceans such as Mysticeti and Odontocet species, dugong (Dugong dugon), Irrawaddy dolphins (Orcaella brevirostris) and Indo-Pacific humpbacked dolphin (Sousa chinensis) are spotted in Brunei Bay. Besides urbanization and industrialization, logging activities and shipping activities have found to be increased around the coast of Brunei Bay. Furthermore, during the field observation, there is also a large scale of paper mill is actively operated on the coast of Brunei Bay.
Surface sediment samples were collected at 16 sampling locations during May, July and October 2013 as well as in January and April 2014 (Fig. 1). Ponar grab was used to collect surface sediment. Plastic scoop was used to sample approximately top 5cm of the sediment. Double layer of polyethylene zip lock bags was used to store the samples. During transportation to the laboratory, samples were kept at low temperature, and later samples were kept frozen in the laboratory. WTW pH meter Multi 340i WTW 82362 Weilheim was used to measure pH of surface sediment in-situ.

Sample analysis
Sediment samples were dried under 60 °C oven heating prior sediment digestion for metals analysis, total organic carbon and particle size analysis. A bulk of 0.05 g dried sediment was digested in the Teflon bomb using a mixture of concentrated acids (nitric acid + hydrochloric acid + hydrofluoric acid) under the oven heating (Ghazali et al., 2016). Digested samples were then diluted and analysed using Inductively Coupled Plasma Mass Spectrometer (ICPMS) ELAN 9000 to determine the concentration of metals in the sediment. The digestion technique was validated against NIST 1646a (Standard Reference Material-Estuarine Sediment, National Institute of Standards and Technology). The concentration of metals in the sediment was calculated using Equation (1)   and the NIST 1646a recovery result is recorded in Table 1. where Y: raw data by ICPMS (µg/L) and Stock volume: top-up volume after digestion process.
The content of total organic carbon (TOC) in the collected sediment samples was determined using method by Walkley & Black, (1934) and Ghazali et al., (2016). The dried sediment samples were digested using water bath heating and titrated. The method was validated using glucose which was 105.5 % recovered value. The percentage of TOC in the sediment sample was calculated using Equation (2): where V1: volume of K 2 Cr 2 O 7 (mL), V2: volume of FeSO 4 (mL) and 0.003: the quantity of carbon in 10 mL of potassium dichromate. Furthermore, the sediment particle size analysis was adopted from Ellis & Stone, (2006) and Ghazali et al., (2016). Dried sediment samples were sieved using a set of a series of eight sieves with mesh sizes of 1 mm, followed by 710 µm, 425 µm, 355 µm, 250 µm, 180 µm, 125 µm and 63 µm. Later, if the weight of particle size under 63 µm is more than 5% out of the total weight, this particular size was analysed using Particle Size Analyser Malvern Mastersizer 2000 and all results were computed using Moment Method.

Data analysis
In the present study, the raw data of metals concentration, percentage of TOC and sediment particle size was computed using DA and PCA in order to further clarify the source of metals in the Brunei Bay surface sediment. Both analyses were carried out using XLSTAT software. DA is a classification type analysis which determines the common properties based on the relationship of a parameter with a particular group. DA creates discriminant function specifically accepting the parameter discriminant ability without calibration (Singh et al., 2005;Kannel et al., 2007;Stella, 2019). The discriminant function was computed using Equation (3): where d i : discriminant function coefficient, Z: the score on each predictor, n: total number of variables.
In the present study, DA was utilized to discriminate the most significant metals which were affected by the spatial and temporal variation based on the variables common attributes. The analysis was applied on the raw dataset and analysed through standard mode, forward stepwise mode and backward stepwise mode. In the forward and backward stepwise modes, the parameter was included and excluded, respectively, until no more significant changes was found (Simeonov et al., 2002).
PCA is a technique used to decrease the multivariate problem dimensions (Helena et al., 2000;Iwamori et al., 2017). In the present study, PCA was    Results and discussion Table 2 shows the variation of metals concentration and the physico-chemical properties in Brunei Bay surface sediment samples, collected throughout the sampling period. Generally, based on the mean value, May 2013 has recorded the lowest concentration of Cu and July 2013 showed the highest concentration of Fe, Pb and pH with the lowest concentration of Al and clay content. Furthermore, the concentration of Al, Cu and sediment particle sizes were highest, plus the concentration of Zn, Cd and TOC were lowest in October 2013. Sediment pH and the concentration of Fe were recorded lowest in January 2014. In the meantime, April 2014 displayed the highest mean value for Zn, Cd and TOC plus lowest mean value of Pb, sand and silt. The overall mean concentration of metals in the surface sediment of Brunei Bay was decreased following the sequence of Al > Fe > Zn > Cu > Pb > Cd. Generally, the overall mean sequence pattern is similar to sequence pattern found during January 2014 and April 2014 sampling period. However, the sequence of mean concentration for Zn, Cu and Pb varied differently during May, July and October 2013. The mean concentration was decreased following the sequence of Zn > Pb > Cu for May and July 2013 as well as Cu > Zn > Pb during October 2013. The change of sequence order for metals in the study area is presumably affected by the seasonal changes. In general, the Brunei Bay is located within the vicinity of the South China Sea; where it receives annual heavy rainfall during the northeast monsoon during early of November to March as well as experiencing the typical dry weather during the southwest monsoon within the period of May to September (Goswami, 2005;Yihui & Chan, 2005;Suhaila & Jemain, 2009). Therefore, the sequence of metals mean concentration in Brunei Bay surface sediment during October 2013 represented the inter-monsoon changes, the period of January 2014 and April 2014 signified the northeast monsoon period whereas the southwest monsoon period was characterized by May 2013 and July 2013.
Based on metal sequence pattern, the mean concentration for metals in Brunei Bay sediment shows the influence of seasonal changes in metals distribution, temporally. The discriminant analysis was utilized on the raw dataset to define which parameters are significant in terms of temporal variation. Table 3 shows the temporal classification matrix of metals in the studied area using standard (A), stepwise forward (B) and stepwise backward (C) DA model.
The discriminant function (DF) for temporal variation of metals using standard mode (Table 3A) shows an accuracy of 85.71% with eight variables. The DA model shows significant variation of data for Al, Zn, Cd, Fe, Pb, pH, sand and clay between all sampling periods. Al, Zn, Cd, Fe, Pb, pH, sand and clay has shown significant variation in between different sampling trips, as shown in Fig. 2. In the meantime, the accuracy of temporal classification of metals using stepwise forward mode is 80% (Table 3B) with five discriminant variables whereas the accuracy of temporal variation of metals via stepwise backward model is 82.86% (Table 3C) with 6 discriminant variables. Based on the stepwise forward model, Al, Zn, Cd, Fe and sediment pH has shown significant changes throughout all sampling periods whereas the stepwise backward model has highlighted Al, Zn, Cd, Fe, sediment pH and clay fraction were significantly varied throughout all sampling periods (Fig. 2) compared to other variables.
The Pearson correlation was used to define the magnitude and direction of the association of two variables. Based on the result obtained, Al showed significant correlation (p < 0.005) with Zn, sand, silt and clay; proclaimed the possible of Zn to be derived by natural sources. This is due to extremely high content of Al is naturally found in the earth crust (Wedepohl, 1995) and sand, silt and clay are the primary texture of soil. However, the significant negative value of R was found between Al and TOC (p < 0.005; R = − 0.4374) signified the anthropogenic sources of TOC as it may be derived by the river discharged surround the bay as most of all ◂ sampling location were located within the estuarine area of the bay system. The finding is similar to those found by Asmala et al., (2019) the upward trend of TOC content in the riverine was corresponded by the anthropogenic discharged into the riverine ecosystem. Furthermore, based on the Pearson correlation, Cd, Pb and Fe was probably derived by the natural sources as Cd was significantly correlated with Zn (p < 0.005); whereas Pb and Fe were significantly correlated with silt but negatively correlated with TOC.
Further clarification on the source of metals was defined via principal component analysis (PCA). Based on the screen plot (Fig. 3), PCA has identified 11 principal components (PCs) which are accounted to affect the distribution of metals throughout the study period. However, only four significant PCs with eigenvalue larger than 1 were rotated using varimax method and highlighted as the sources of metals in Brunei Bay surface sediment.
The four significant PCs have contributed approximately 77.31% towards the distribution of metals in the Brunei Bay (Table 4). The first PCs with 38.62% of contribution have classified Al and sediment type (sand, silt and clay) into one significant PC, and it was proven that contribution by the earth crust is dominating the distribution of metals in Brunei Bay's surface sediment. This is due to Al is the most abundance metals exist in the earth crust, approximately 8% (Wedepohl, 1995) and owned the strongest relationship with sand, silt and clay content the study area. In numerous studies elsewhere, Al has shown strong correlation with soil and sediment due to high concentration in the earth crust, naturally (Angelova et al., 2020;Haider et al., 2020).
Furthermore, Zn and Cd have shown strong factor loadings on PC2 with 15.2% out of the total   variance, whereas Fe with 14.2% of variance and Cu with 9.3% of variance has strong factor loading on PC3 and PC4, respectively (Table 4 and Fig. 4). Based on the factor loading plot (Fig. 4), these metals are prone to derived by natural sources but most probably also contributed by some anthropogenic input, as they are mostly distributed in the same axis with the sediment characteristics. Generally, Fe, Cu, Zn and trace content of Cd are knowingly essential elements in living organisms and they naturally exist in the environment. These elements are widely supplied and used in various activities such as in farming, industrialization, aqua-farming, urbanization and many more (Nazir et al., 2015;Xiao et al., 2019a;Wang et al., 2019). While doing so, some of the surpluses are excessively flushed out and end up in the sediment bed. The Brunei Bay vicinity is collectively occupied by local activities such as small scale of farm fishing, boating, local fishing, agriculture as well as providing shelter for ships during the monsoon season. However, based on chemometrics interpretation, such activities did not affect much on the bay's ecosystem since the bay ecosystem still can provide a vast array of biological diversity (Jaaman et al., 2008;Satyanarayana et al., 2018). This finding agrees to the past findings (Ghazali et al., 2016;Adiana et al., 2017) as proclaimed that metals distribution in the surface sediment of Brunei Bay was dominantly influenced by the natural input. Nevertheless, the increasing of human activities to fulfill their living need is slowly contributing slight anthropogenic metals into the bay's ecosystem, gradually.

Conclusion
The current study has utilized the chemometrics analysis on the secondary data to further clarify factors which has been affecting the distribution of metals in Brunei Bay's sediment. Based on the mean value, the metals sequence changed temporally due to the changes of season. Moreover, the DA has showed that the seasonal changes were significantly affecting the distribution of Al, Zn, Cd, Fe, sediment pH and clay fraction in the bay's sediment. Besides that, the Pearson correlation and PCA verified the natural sources of metals were dominated the distribution throughout the study period, whereas the anthropogenic activities were lightly proven to influence the metals distribution in the bay's sediment. Likewise, the increasing of human activities within and surround the bay area are prone to worsen the bay's ecosystem health in time.