The 110 surface sediment samples analyzed in this study were collected from the mid-Bay of Bengal between March and May 2014 by The First Institute of Oceanography (FIO), State Oceanic Administration (SOA), China using a box corer (size: 30 cm×30 cm×65 cm; a mass: 200 kg; sediment: approximately 70 kg). This corer had a small volume, was easy to operate and caused minimal disturbance, therefore, it is ideal for stratified sampling. Light yellow-colored fluid mud was observed on the surface of the samples and suggested that the surface sediments were well preserved. Sediments collected in the upper 5 cm were stored and analyzed to ensure isochronism of the samples.

The measurements of the grain size distributions of carbonate-free terrigenous particles were carried out using a laser grain-size analyzer (Malvern 2000, range: 0.02–2 000 μm, resolution: 0.01Φ). Prior to the analyses, 15 mL of 3% H₂O₂ and 5 mL of 3 mol/L hydrochloric acid (HCl) were added to remove organic matter and carbonate fractions, respectively. Then, the mixture was rinsed several times using deionized water and extracted via the centrifugation. After ultrasonic oscillation, the samples were measured. The relative error of repeated measurement was less than 3%. Preparation and measurement were completed at the Key Laboratory of Marine Sedimentology and Environmental Geology (MASEG), FIO, SOA, China.

The freeze-dried samples were powdered to 200 mesh and dried. Each sample (0.05 g) was weighed, placed in a polytetrafluoroethylene digestion tank, dissolved twice in HF-HNO₃ (1:1), and dried again at 190°C for 48 h. 3 mL of 50% HNO₃ was then added to each sample, and the samples were dried at 150°C for at least 8 h, removed and analyzed. Element concentrations were analyzed using an inductively coupled plasma optical emission spectrometry (ICP-OES; Si, Al, Ca, Fe, K, Mg, Mn, Na, P, Ti, Ba, Sr, V, Zn, and Zr) and inductively coupled plasma-mass spectrometry (ICP-MS; Sc, Cr, Co, Ni, and Cu). Several samples were analyzed in replicate to determine the precision of the measurements, and the elemental compositions of a GSD-9 reference standard were measured to confirm the accuracy of the analyses, which was greater than 2% for most elements. The analyses were also performed at the Key Laboratory of MASEG, FIO, SOA, China.

The mean grain size (Mz) varies between 6.2Φ and 7.6Φ, with an average of 7.1Φ. Fine grain fractions were mainly distributed in the northern and southern parts of the area, with a size range of 7.1Φ–7.6Φ, while the coarse fraction was mainly distributed in the central part of the area, with an Mz below 7.1Φ (Fig. 2a). Four types of sediments were observed in the study area: silt, sandy silt, mud and sandy mud (Fig. 2b).

The concentrations of major and trace elements in the surface sediments are summarized in Table 1. Among the analyzed major elements, Si exhibited the maximum average concentration (18.80%) and P exhibited the minimum average concentration (0.04%). Among the analyzed trace metals, Ba exhibited the maximum average concentration (834.42 μg/g), and Sc exhibited the minimum average concentration (15.93 μg/g). Isoconcentration maps of selected typical elements are presented in Fig. 3. The distribution features of Si, Al, and Ti were similar, with higher concentrations in the southern and northern areas of the study area and lower concentrations in the northwestern, northeastern and central areas of the study area. By contrast, Ca concentrations were higher in the northwestern, northeastern and central areas. The distribution features of Cu and Co indicated relative homogenization, with no apparent higher and lower areas.

The chemical composition of marine sediments is determined by the proportions of materials derived from detrital, authigenic and hydrothermal sources (Tripathy et al., 2014). Detrital material and biological components play dominant roles in the study area (Kessarkar et al., 2005; Pattan et al., 2008). To determine the relative contributions of different provenances, the different source materials must have different chemical compositions, and the source signatures must be well preserved in the sediments of the study area. The chemical composition of river sediments is likely to remain mostly unaltered in the sea (Tripathy et al., 2014), and consequently, the comparison of the chemical composition of river sediments with that of marine sediments can provide useful information regarding provenance contributions.

The chemical index of alteration (CIA) can provide useful information about chemical weathering. The extent of weathering loss for detrital components can be estimated by comparing the CIA values of sediments in the study area with those of the source materials. Here, the modified CIA* (no CaO included, the oxide abundances are in molar units) (Tripathy et al., 2014) is expressed as follows:

The CIA* values for major rivers in the potential provenances are presented in Table 1. The CIA* for the surface sediments of the study area is intermediate between the CIA* for sediments from the Himalayan and Indian rivers, with an average CIA* value of approximately 72.07. This results confirms that the composition of the source material is preserved in the Bay of Bengal sediments and therefore can be used to trace their provenances and estimate relative contributions to the study area.

The factor analyses of the geochemistry data set were performed using IBM SPSS Statistics 19.0 software to further constrain the factors influencing the geochemistry data for the surface sediments in the study area (Table 2). The following conditions must be satisfied in the factor analyses: (1) sufficient number of samples, generally, the number of samples must be at least five times the number of variables; (2) a strong correlation must exist between the original data, the factor analyses can be performed when the Bartlett test of sphericity yields a value under 0.05 or the Kaiser-Meyer-Oklin (KMO) measure of sampling adequacy is above 0.5; and (3) the common factors obtained must be significant, and a factor rotation can be used if necessary. In our study, 110 samples were analyzed, and the number of samples was five times large than the number of variables. The Bartlett test of sphericity and the KMO measure of sampling adequacy were 0 and 0.85, respectively. Thus, it was appropriate to perform the factor analyses. An oblique rotation, which requires oblique relationships between factors, was used. Three major factors with eigen values greater than 1 that together explain approximately 90% of the variance of the data set were extracted. The dominant factor (Factor 1 (F1)) explains approximately 58% of the total variance and is loaded primarily with lithogenic elements, e.g., Si, Al, Ti and Fe. Ti and Al frequently reflect the detrital component. These elements tend not to migrate in weathering processes and can preserve the source signatures; therefore, they are assigned as the detrital source in Factor 1. Factor 1 is also characterized by higher loading with various trace elements (V, Cr, Zr, and Sc, Table 2), which implies that their dominant source is also terrigenous. The higher loading of Fe and weak loading of Mn (much less than in Factor 2 (F2)) in Factor 1 and the weak correlation (0.40, Table 3) between Fe and Mn indicate that Fe in these sediments is primarily of detrital origin but is minimally related to Fe-Mn oxide. The typical biophile elements, Ca, Ba, and Sr, exhibit higher negative loadings in Factor 1 because their concentrations are primarily controlled by carbonate and biological deposition, which were eliminated in the pre-treatment process. The element associations in Factor 2 are Mn, Na, Zn, and Ni. Generally, Mn in marine sediments is derived from authigenic and detrital sources (Shi, 2012); Na is a mobile element and also tends to concentrate in fine-grained sediments as a result of adsorption and cation exchange between water and sediments (Zhao and Yan, 1994). Mn and Na exhibit no apparent correlation with detrital indicator elements (e.g., Al and Ti) (Table 3); therefore, they were assigned as the authigenic deposit in Factor 2. Co and Cu, which are essential elements for phytoplankton, are controlled by Factor 3 (F3). They can be absorbed by phytoplankton and deposited with its debris. The contents of Co and Cu in sediments are closely associated with organic matter and a primary productivity (Yu et al., 2012); therefore, Factor 3 represents the effect of the primary productivity on elemental characteristics.

The discriminant function (DF, $\,f_{\rm d}$) describes the relationship between sediments in the adjacent areas (Lan et al., 2009) and is widely used for provenance discrimination. The DF is calculated as follows:

where ${c_{\rm 1x}}/{c_{\rm 2x}}$ represents the concentration ratio of two elements in the sediments in the study area; and c_1L/c_2L is the concentration ratio of two elements in the sediments in the adjacent area. Small DF values (generally below 0.5) indicate close associations between the chemical components in sediments of the study area and chemical components in sediments of the adjacent area (Li and Li, 2001). The ratios of elements to Al can eliminate the effects of the grain size variation on the element components because Al remains stable under changing environmental conditions (Kremling and Streu, 1993). Correlation coefficients between Mz and the ratios of elements to Al were calculated, and the results imply weak correlations from 0 to 0.42. The value of Mz and the ratios of the seven primary controlling elements in Factor 1 (Mg, Ti, Fe, V, Cr, K, and Zn) to Al exhibit the weaker correlations (0–0.26); thus, they were used to calculate the DF. The DF between the study area and the Himalayan source was lowest, with a value of 0.15, followed by the Indian source, with a DF of 0.24; the maximum DF value, 0.51, was observed for Myanmar. On the basis of the DF results, the primary provenances for sediments in the study area are Himalayan and peninsular India, with little contribution from Myanmar.

Owing to the dilution of terrigenous and biological (mainly from ocean processes) material (Xie et al., 2014) and the effect of the grain-size variation (Kremling and Streu, 1993), a single element index cannot accurately reflect the source, and choosing appropriate ratios between elements can eliminate these effects on the two elements simultaneously. Ti is also used to indicate the input of terrigenous material (Taylor and McLennan, 1985; Yarincik et al., 2000; Lebreiro et al., 2009). Because Ti is generally found in rocks of continental crust origin (Taylor and McLennan, 1985), is not sensitive to oxidation-reduction and is essentially unaffected by diagenesis, Ti-normalized element ratios are useful for sediments (Xie et al., 2014). As shown in Table 3 and Fig. 4, strong correlations between Ti and several elements were observed (r²>0.7). Several elements (e.g., Al, Fe, Mg, and Cr) are supplied to the Bay of Bengal from continental aluminosilicate (Sarin et al., 1979), but the correlation among Fe, Cr, and Ti is closely related to the presence of Fe-Ti minerals from the Himalayas (Garzanti et al., 2011) and Deccan Traps (Das and Krishnaswami, 2007) and is not affected by weathering and the interchange between Cr³⁺ and Fe³⁺ in the mineral structure (Schwertmann and Cornell, 2000). This result indicates that the element concentrations in the surface sediments in the study area are dominated by terrigenous detrital material and, consequently, can provide information about sources.

A geochemical data set inverse model (Tripathy and Singh, 2010; Tripathy et al., 2014) was used in this study to quantify the contribution from major provenances. This model includes a set of mass balance equations, beginning with pre-assigned elemental ratios for the sources, and evaluates the relative contributions of each of the sources to the study area (Tripathy et al., 2014). In this work, the model is used to apportion the contributions from two major sources: Himalayan (represented by G-B) and peninsular Indian (represented by G-K). Elements with high factor loading in Factor 1 (Mg, Ti, Al, Fe, V, Cr, and K) were used in the mass-balance equation (Table 2).

The equation for an element (X) in sediments of the study area can be written as follows:

where X_i represents the element concentration for a given source i (i=1, G-B; i=2, G-K); f_i represents the contribution of sediments from source i to the study area; and n is the number of sources, which was 2 in this work.

For Ti-normalized ratios of the element (X), Eq. (3) can be rewritten as follows:

Equations (3), (4) and (5) for elements X=Mg, Al, Fe, V, Cr and K were used to estimate the source contributions. The element concentrations for sediments with Himalayan and Indian sources are presented in Table 1. The inverse model results are presented in Table 4. The Himalayan source supplies approximately 83.5% of the sediments to the study area, and approximately 16.5% of the sediments are derived from the Indian source.

On the basis of the model results, the Himalayan source plays a major role in the contribution of sediment, with minor contributions of the Indian source. The distributions of proportions of contribution from the Indian source and the Himalayan source are presented in Fig. 5. To better illustrate the relative contributions in different areas and considering the relatively low values of Indian provenance, a 10% interval was used. The sites with a larger proportion of sediments derived from the Indian source are mainly located in the western part of the study area, closer to the Indian continent, with contributions of Indian material mostly greater than 10%. The eastern part of the study area, which is further from the Indian continent, receives less from the Indian source than the western part of the study area, with contributions of Indian material of mostly less than 10%. The eastern area received more material from the Himalayan source than the western area.

During sea-level high stands, such as in the present time period, the main path for sediments to the Bengal Fan is the “Swatch of No Ground”, which is along a submarine canyon located in the slope of the northern part of the Bay of Bengal. The fan does not receive a direct load from the G-B rivers but does intercept sediment moving on the continental shelf (Weber et al., 1997). Slump activity in the existing submarine canyon is weak; hence, turbidity currents are weaker than during sea-level low stands and occur mostly in the upper and middle fan area (Curray et al., 2002). The currents are constrained within the large channels, and less overflow occurs. Although turbidity becomes weaker in channels from the upper fan to the middle fan, the channel area in the middle fan becomes much smaller compared with the upper fan; hence, overflow occurs, and the suspended sediments carried in this channel area are deposited as the strength of the turbidity currents decreases. By contrast, the peak input of terrigenous flux derived from the G-B rivers into the bay is during SW monsoons (Goodbred Jr, 2003; Mergulhao et al., 2013; Tripathy et al., 2014). The plume of fresh water extends southwards up to 15°N (Chauhan and Vogelsang, 2006), and suspended sediments deposit slowly during dispersion. These two mechanisms create the primary patterns for supplying sediments from Himalayan sources during the present sea-level high stands.

Submarine canyons in the slope of the western part of the bay are small in scale; therefore, there is no submarine canyon similar to the “Swath of No Ground” to transport sediments from the Indian source to the Bengal Fan. The flux of Indian rivers and their sediment flux to the bay are approximately 5 times and 4 times smaller than the G-B, respectively (Milliman and Farnsworth, 2011); therefore, the plume of fresh water has a smaller range, and surface circulation plays a more important role in the transport of the Indian source material to the study area. The transport of Indian material to the northeast is accomplished through the action of anticyclonic currents, particularly the East India Coastal Current during the SW monsoon, which carries material from the Godavari-Krishna and Mahanadi rivers and appears to affect the western part of the study area (Shankar et al., 2002; Narvekar and Kumar, 2006). Therefore, the western part of the study area receives sediments more readily from the Indian source and is affected more by this deposition than the eastern part of the study area.