Sulfur and molybdenum fractionation in marine and riverine alluvium paddy soils

Abstract Intermittently submergence and drainage status of paddy fields can cause alterations in morphological and chemical characteristics of soils. We conducted a sequential fractionation study to provide an insight into solubility of Sulfur (S) and Molybdenum (Mo) in flooded alluvial paddy soils. The samples (0–15 and 15–30 cm) were taken from marine and riverine alluvial soils in Kedah and Kelantan areas, respectively, and were sequentially extracted with NaHCO3, NaOH, HCl, and HClO4–HNO3. Total S in upper and lower layers of Kedah and Kelantan ranged between 273 and 1121 mg kg−1, and 177 to 1509 mg kg−1, respectively. In upper layers and subsoil of Kedah, average total Mo were 0.34 and 0.27 mg kg−1, respectively. Average total Mo in Kelantan were 0.25 mg kg−1 (surface layer) and 0.28 mg kg−1 (subsoil). Cation exchange capacity (CEC) was positively correlated with plant available amounts of Mo in upper layers of Kedah area. Also, total and medium-term plant-available S was correlated with total carbon (C) at lower layers of Kelantan soil series. But in surface layers of Kelantan soil series, CEC was strongly correlated with total and medium-term plant-available S. Our results indicates that the influence of flooding conditions on soil S and Mo contents in paddy fields may cause long-term changes in S and Mo chemical reactivities.


Introduction
Flooding soils with water causes drastic changes in normal physical, chemical and biological processes, and reactions. The heterogeneity of these soils, in terms of fluctuations in redox potential and their chemical specificities, are result of changes in distribution of microorganisms, mineral phases, and mixing different redox reaction substrates and products. [1] Rice, as a traditional staple food in Asia, is grown as wet land rice and is arguably the most important crop in the world. In peninsular Malaysia, rice-growing areas are categorized into three groups as follows: (1) coastal alluvial plains that are mainly found on the west coast; (2) river terraces of the large rivers which are mainly seen on the east coast; (3) floodplains and valleys of the small rivers. [2] Sulfur, as an important element in paddy fields, undergoes redox transformations. In oxidizing conditions, sulfate ions (SO −2 4 ) can move easily with soil solution and in this form they are taken up by plants.
In acidic soils, if microbial activities are reduced, plant available form of sulfur will be decreased due to low S oxidation. [3] It is generally stated that pH values, type of cation present, presence of competing anions, extractable amounts of Al 3+ and Fe 3+ , organic C, clay content, and soil horizon type govern to a vast extent sulfate sorption in various soils. [4] Aluminium-and Fe-humus complexes and inhibition of crystallization of Al and Fe oxides by organic matter are accounted for increased reactive surfaces toward sulfate adsorption. [5] The initial increase in water-soluble sulfate at the beginning of submergence and then a slight decrease for several months is seen in acid soils. [6] It was reported that some coastal sediments in west coast of peninsular Malaysia contain up to 2-3% pyrite as a result of thousands years of pyritisation process. [7] Pyrite (cubic FeS 2 ) is the most thermodynamically stable iron sulfide mineral in soils [8] which is produced by transformations of unstable monosulfides. [9] When sulfate is reduced, hydrogen sulfides (H 2 S), which are second mechanism for pyrite formation, are formed. Prevalence of this gas is only seen in sandy soils low in reducible iron or the soils contain large amounts of organic matter under flooding conditions.[1] In the presence of H 2 S, thiooxymolybdates such as MoO 6 S 2− , MoO 3 S −2 , and MoO 2 S −2 2 are formed and MoO 2 S −2 2 becomes dominant when concentration of sulfide enhances. [10] According to Wang et al. [11], The presence of relatively stable residual Mo (VI) under high concentrations of H 2 S supports and riverine alluvial materials from Kedah and Kelantan states, Malaysia.
The average annual rainfall in Kedah and Kelantan is 2201.9 and 2854.4 mm, respectively. [21] In Kedah and Kelantan, the lands under rice cultivation were about 50.8 and 10.5% of their whole land areas, respectively. [22] Also, the mean temperatures were not so different between Kedah (27.6 °C) and Kelantan (27.4 °C) states. [21] Rice harvested area in Kedah (187.413 ha) was approximately 5 times larger Kelantan (38.641 ha). [22] We collected different samples of soils from two above-mentioned areas on paddy fields (Table 1). [23]

Determination of soil properties
The integrated soil samples were collected from depths of 0-15 and 15-30 cm. For analysis, soil samples were oven-dried and visible macro fauna and root pieces were removed. Samples were crushed and sieved to <2 mm. Silt and clay concentrations were determined by pipette method, and fine and coarse sand particles were separated by sieving (50-250 and 250-2000 μm diameter). Soil pH was measured by a glass electrode in distilled water and 0.01 M CaCl 2 at a soil-to-solution ratio of 1:2.5 (w/v).
Soil samples (0.5 g) were accurately weighed and placed in ceramic crucibles along with combustion accelerators. The crucibles were placed in a high temperature furnace which was then flooded with oxygen (Trumac Determinator, leco USA). The furnace was heated to 1350 °C and caused the combustion of the carbon and sulfur in soil samples. Cation exchange capacity (CEC) was calculated by sum of cations using 1 M NH 4 OAc (pH 7.0). Amorphous iron (Fe ox) and aluminum (Al ox) were determined by extraction in solution using oxalic acid and ammonium oxalate (pH 3) for 4 h in the dark, [24] following by detection using inductively-coupled plasma optical-emission spectrometry (ICP-OES) (Optima 8300, PerkinElmer, USA). The samples high in clay contents were centrifuged at high speeds. Total molybdenum (Mo) was determined independently of fractionation procedure by solubilizing Mo with HCl at 120 °C and use of heptan-2-one as a solvent with detection by ICP-OES. [25] the assumption that thiomolybdates are of diagenetically importance in sulfidic sediments.
However, mineralization of organic sulfur compounds or immobilization of them into organic compounds are seen in oxidized and reduced zones of waterlogged soils but production of sulfide prevails mainly in reduced soil layers. [12] Molybdenum occurs in different potential soil fractions including positively charged surfaces, crystalline material, soil solution, organic material, and discrete compounds. [13] Molibdenite (MoS 2 ) is the most abundant form of Mo in igneous rocks. In general, rates of 0.5-5 mg kg −1 are considered normal, and soils contain 0.5 or less could be considered low. [14] Prevalence of Mo in soil in different oxidation states (2)(3)(4)(5)(6) have made the chemistry of Mo more complex. lindsay, [15]  . Molybdenum could be traced in some fractions such as soil solution, organic matter, and solid phases and are mainly bound with oxides. Inner and outer -sphere surface complexes for Mo adsorption were found on Fe and Al oxides, respectively, but MoO 2− 4 forms innersphere surface complexes on both Al and Fe oxides at low pH and outer-sphere surface complexes at high pH. Adsorption of Mo on a weight basis on all soil Fe and Al oxides is higher than adsorption on clay minerals, and its adsorption on montmorillonite is higher than illite and kaolinite. [16] McGrath et al. [17] also stated that molybdate sorption on amorphous iron oxides largely affects Mo toxicity under different pH regimes. It was also reported that retention of molybdate on γAl 2 O 3 was higher than sulfate, and molybdate considerably inhibited adsorption of sulfate on aluminum oxides but sulfate did not have any effect on molybdate adsorption. [18] During oxic conditions, molybdenum exists in oxidation state of 6 in oxyanionic form but in anoxic conditions, sulfide mineral molybdenite is predominant. [19] Under water-logged conditions, two layers are produced. The surface layer contains high levels of dissolved oxygen, whereas oxidized oxygen is not present in underlying layers. Hence, redox potential is low at surface layers and increases with depth. Fox and Doner, [20] reported that maximum accumulation of Mo was found near water-sediment interface which might resulted from formation of MoS 2 and MoS 2− 4 . Case studies with a focus on total molybdenum status in soils have been rarely done in Malaysia and this is the first study, to our knowledge, to test molybdenum concentrations in different fractions of alluvial paddy soils of two major rice production areas in this country. This study was conducted to determine S and Mo concentrations in different rice soils developed on marine  Amorphous iron varied from 0.53 to 2.84 g Fe kg −1 (Kedah), and 0.04 to 1.05 g Fe kg −1 (Kelantan), and total amorphous aluminum between 0.14 to 0.56 g Al kg −1 (Kedah) and 0.08 to 0.55 g Al kg −1 (Kelantan). In Kedah soil series, CEC values varied from 13.26 to 20.85 meq 100 g −1 (0-15 cm) and (9.36) to 27.31 meq 100 g −1 (15-30 cm) and increased with depth except in Guar and Keranji series. CEC of Kelantan soil series ranged from 2.9 to 13.49 meq 100 g −1 (0-15 cm) and 2.67 to 9.13 meq 100 g −1 (15-30 cm) and declined with depth except in lating b series. Average clays percentage of the soils in Kedah found to be higher than Kelantan soil series at both depths. Sedu series had highest clay content (61%) while the lowest was seen in lating a series in the both layers. There was no significant difference in values of soil pH, CEC, amorphous iron and amorphous aluminum between soil depths except in total C contents of Kelantan soil series (*p < 0.05).

Concentrations of sulfur fractions
In this experiment, total sulfur and portions of sulfur are illustrated in Tables 4 and 5 and Figures 2 and 3.

Sulfur and molybdenum fractionation
Triplicate 1 g soil samples were placed into 50 ml centrifuge plastic tubes and sequentially extracted according to a modified version of sequential fractionation procedure by Hedley et al. [26], represented by lilienfein et al. [27]. This sequential extraction method was used to assess S and Mo concentrations because these elements are of certain general similarity in chemical characteristics. labile or weakly adsorbed S and Mo were extracted using 0.5 M sodium bicarbonate (NaHCO 3 ). After 16 h of shaking (end-overend), supernatants were centrifuged and filtered. Bounded S and Mo to Al and Fe hydrous oxides (medium term plant available S and Mo) were determined by extracting the soil with 0.1 M NaOH. Samples were shaken for 16 h, followed by centrifuging and filtering. Acid soluble S and Mo (associated S and Mo with calcium and iron oxides) were extracted using 1 M HCl, and after 16 h supernatants were centrifuged and filtered. Residual fraction (highly resistant and occluded S and Mo) were digested by application of 5 M HNO 3 and concentrated HClO 4 at 200 °C. After filtering, all aliquots were analyzed by ICP-OES.

Data analysis
Sulfur and Mo are demonstrated as the mean of the three replicate extracts. Correlations between soil properties and S and Mo fractions were determined using SAS 9.4, Sigma plot version 12.0 and Excel softwares. Significant differences between soil depths were calculated using Students's t-test at p < 0.05. All results were stated on the basis of oven dried soils at 105 °C for 24 h.

Soil properties
Chemical and physical properties of soils are illustrated in Tables 2 and 3. In upper layers of Kedah and Kelantan, contents of clay varied from 39.5 to 61.2% and 5.8 to respectively ( Figure 2). In subsoils of Kedah soil series, NaHCO 3 and HCl extracted the largest and smallest S fractions, respectively ( Figure 3). NaHCO 3 -extractable S accounted for between 31.3 and 56.7% of the total S in lower layers of Kedah soil series, respectively ( Figure 3).
In surface layers of marine alluvium soil series (Kedah area), NaOH and HCl fractions extracted the largest and smallest average S contents, respectively ( Figure 2). Sulfur in NaOH and HCl extracts accounted for between 10.6 to 49.2% and 1.9 to 18.9% of the total S in higher layers of marine alluvium soil series,  area, proportions of S declined along the line HClO 4 -HNO 3 > NaOH > NaHCO 3 > HCl ( Figure 5). Cempaka (Bukit Makrifah) soil series contained the highest NaHCO 3 -extractable S (250.05 mg S kg −1 ) in surface layers but the largest amounts of S, which were extracted by NaOH, HCl, and HClO 4 -HNO 3 , were seen in Batu Hitam soil series (Table 5). Also, the lowest amounts of extracted S in HCl (8.6 mg S kg −1 ), NaHCO 3 (43.05 mg S kg −1 ), and NaOH (60.8 mg S kg −1 ) extracts were seen in upper layers of lating B series (Table 5).

Concentrations of molybdenum fractions
Molybdenum concentrations in sequential extractions are shown in Figures 6-9. In Kedah area, most of Mo levels were extracted with NaHCO 3 and NaOH fractions at both depths. Molybdenum in NaHCO 3 extracts accounted for 16.5-83.7% of total Mo at upper layers of marine alluvium soil series (Kedah area) ( Figure 6). Mean concentrations of Mo extracted by NaHCO 3 and NaOH Also, greatest amounts of extracted S were observed in NaHCO 3 (730.5 mg S kg −1 ) and NaOH (401.6 mg S kg −1 ) fractions in subsoils of Sedu series ( Table 4).
The maximum level of S (850.45 mg S kg −1 ) was extracted by HClO 4 -HNO 3 in surface layers of Kundur series, while the lowest amount of S (19.7 mg S kg −1 ) was extracted by HCl in Sedu series of Kedah Area (Table 4). In subsoils of Kedah soil series, amounts of extracted S in a descending order were as follows: NaHCO 3 > NaOH > HClO 4 -HNO 3 > HCl (Figure 3).
According to the results, the proportions of S decreased along the line HClO 4 -HNO 3 > NaOH > NaHCO 3 > HCl in surface layers of riverine alluvium soil series (Kelantan aera) (Figure 4    According to results, Mo concentrations in NaHCO 3 extracts were not detected in lower depths of Keranji and Rotan soils series (Table 6) whereas, 98 and 95% of total Mo were obtained in NaOH extract in subsoils of Keranji and Rotan series, respectively (Figure 7). The highest amount of Mo in NaHCO 3 extracts was achieved in subsoil of Guar series (0.28 mg Mo kg −1 ) ( Table 6).
Rotan soils contained the highest NaOH-extractable Mo (0.35 mg Mo kg −1 ) in lower depths (Table 6). We did not detect any Mo in HCl extracts at both depths of Kedah soils (Table 6). In Kedah, residual Mo of upper layers was found in different soils but in Guar, Kangkong and Rotan series ( Table 6). In lower layers, residual Mo was not detected in any soils except in Sedu series (0.04 mg Mo kg −1 ) ( Table 6). In riverine alluvium soil series (Kelantan area), mean concentrations of NaHCO 3 -extractable Mo in subsoils (0.25 mg Mo kg −1 ) were nearly two times higher than surface layers (0.11 mg Mo kg −1 ) ( Table 7).
In subsoils of Kelantan area, proportions of Mo decreased along the line NaHCO 3 > HCl > NaOH > HClO 4 -HNO 3 . The results show that maximum Mo concentration (0.37 mg Mo kg −1 ) was achieved in NaHCO 3 extracts of lower depths of lating a soil series (Table 7). Also, NaHCO 3 extracted 99, 98, 96, and 93% of Mo from subsoils of Cempaka AB, lating b, Cempaka BM, and Cempaka SK series, respectively ( Figure 9). In upper layers, NaOH extracted Mo of all soils, but we did not detect any Mo in NaOH fractions of subsoils except in Batu Hitam series (0.07 mg Mo kg −1 ) ( Table 7). At both layers, Mo was not obtained in HCl fraction except in lating a (both layers) and Batu Hitam series (subsoil). The proportions of Mo declined along the line NaOH > NaHCO 3 > HClO 4 -HNO 3 > HCl in surface layers of Kelantan soil series ( Figure 8). In lower layers, residual Mo was only seen in Cempaka BM, but in surface layers was found in Cempaka BM, lating a, Batu Hitam, and lating b series. Results showed that there was no correlation between total Mo and soil properties. Also, CEC was linearly correlated with NaHCO 3 -extractable Mo (r 2 = 0.61; p < 0.05) at upper layers of Kedah soil series.

Principal component analysis of sulfur
Principal component analysis (PCA) was conducted on the data of soil properties and extracted S and Mo were greater at surface soils than subsoils in Kedah (Table  6) but they were not significantly different. Although Mo was not detected in NaOH extract of Kundur series (subsoil), but NaOH-extractable Mo accounted for between 16.1 and 98% of total Mo at lower layers of Kedah soil series (Figure 7).  has high positive loading on the variables HCl-S and NaOH-S and high negative loadings on Fe ox and clay. In lower layers of Kelantan soil series, PCA results illustrated that three factors all had eigenvalues higher than one ( Table 9). The three principal components explain 97% of total data-set variance. The first principal component explains about 61.5% of the total variance, the second factor explains about 27.03%, and the third principal component explains about 8.9%. The first eigenvector has high positive loadings on variable total S, carbon and NaOH-S and high negative loadings on variable pH. According to this result, carbon, NaOH-S and total S were close together and pH seems to be a non-effective factor, affecting availability of S in lower layers of riverine alluvium soil series. The second eigenvector has high positive loadings on the variables Fe ox and clay and high negative loadings on variables HCl-S-S. The third eigenvalue has high positive loading on the variables pH and high negative loadings on Al ox. in order to better understand their interrelationships. In surface layers of Kelantan soil series, the results showed that three factors all had eigenvalues higher than one so that 94% of total variance with the first three principal components can be explained ( Table  8). The first principal component explains about 58.6% of the total variance, the second factor explains about 19.2%, and the third factor explains about 16.7%. The first eigenvector has high positive loadings on variable CEC, total S and NaOH-S and high negative loadings on variable pH. According to this, CEC, total S and medium term plant available S were fairly close together, representing a close relationship among them. Furthermore, the pH appears to be a noneffective factor, influencing plant available S in surface layers of riverine alluvium soil series. The second eigenvector has high positive loadings on the variables NaHCO3-S and pH and high negative loadings on variables HClO4-HNO3-S. The third eigenvalue   Table 10 illustrates the factors, eigenvalues, and associated variance explained in the physico-chemical data and communities. According to results, three factors all had eigenvalues higher than one (the most significant one). Also, three extracted factors explain 85.7% of data-set variance. The first principal component explains about 38.8% of the total variance, the second factor explains about 33.3%, and the third factor explains about 13.5%. The first eigenvector has high positive loadings on variable NaHCO 3 -Mo, CEC and total carbon, and high negative loadings on variable NaOH-Mo. The second eigenvector has high positive loadings on the variables  anaerobic decomposition in brackish marshes, resulting in regulation of carbon reservoir of soil. This means that large contents of sulfur in the form of sulfate in reducing layers can be directly related to carbon contents in lower reduced layer of flooded soils. Our findings showed that average portion of NaHCO 3extractable S was higher in surface layers than lower depths of Kelantan soils. It has demonstrated that high concentrations of sulfate in oxic conditions at higher E h zones of paddy fields can be attributed to activities of sulfur-solubilizing bacteria, while formation of sulfide at lower E h in paddy subsoils might be due to sulfate-reducing bacteria function. [41] The current results show that as CEC reached the highest value and maximum carbon content was achieved in surface layers of Kelantan (Batu Hiam series), the largest amount of S (352 mg kg −1 ) was also recorded. It demonstrates that medium-term plant available S contents, which are adsorbed more strongly on iron and aluminum hydrous oxides, were increased as capacity of cation exchange and organic content of flooded soils in surface layers were enhanced. In subsoils of Kelantan, concentrations of NaOH-extractable S were high as carbon content increased. Although amorphous iron contents were not correlated with S concentrations in NaOH fractions in lower layers of Kelantan soil series, but their greater levels might have affected the sorption of S concentrations in NaOH fractions of Cempaka series. Recalcitrant S concentrations in HClO 4 -HNO 3 fraction of subsoils were higher than concentrations of recalcitrant S of upper layers of Kelantan soil series. Theses soils are slightly light textured and vertical or lateral soil solution movements may account for sulfur losses in such soils. Furthermore, reduction of ferric forms of iron to ferrous forms and formation of amorphous iron from ferrous iron in lower depths of flooded soils lead in increases of anion sorption capacity and immobilization of S. If the contents of other competing anions for the sorbing surfaces, such as phosphate, are large in reduction layers of soils, the concentration of sulfate in solution will be increased. [42] It was shown that organic Al and crystalline Fe contents accounted for the sorption of sulfate in some spodosols, [43] but in ultisols, retention of sulfate was associated to only amorphous iron not amorphous Al. [44] In contrast to ultisols, retention of sulfate was positively correlated with inorganic amorphous Al in podzolic soils. [45] According to current results, the average total amounts of S declined with depth of soil following same pattern as average total carbon in all soil series. Our results are in agreement with those of Möller et al. [46] who presented that status of S in tropical soils depended closely on the turnover rate and content of organic matter. Also, a positive relationship between organic carbon of soils and water soluble sulfate in forested soils was observed by Neary et al. [47].

PCA of molybdenum
In contrast to Kedah, Kelantan soils series contained higher available S in NaHCO 3 fraction in surface layers

Discussion
In this study, the paddy soils of peninsular Malaysia were found to contain low levels of Mo (0.41-0.87 mg kg −1 ) compared with high Mo contents (6-8 mg kg −1 ) derived from swampy lowland paddy fields of east Malaysia. [28] Concentration of Mo in alluvial soils, spodosols, and inceptisols are quit low, and are generally ranged between 0.2 and 5 mg kg −1 in soils. [29] The current results illustrated that CEC was positively correlated with NaHCO 3 -extractable Mo in surface layers of Kedah soil series. Molybdate (MoO −2 4 ) is the oxyanion which is commonly found in solutions under oxidizing conditions. In a laboratory study, Mo was immobilized under reducing conditions and remobilized under oxic conditions, and iron minerals became major sinks for Mo adsorption in reducing conditions. [30] Also, CEC enhancement coincided progressively with decline in E h in the AP horizon of a waterlogged rice field. [31] Redox potential is high in surface layers of flooded soils and decreases with depth to low negative values. [32] According to our results, Keranji series had the highest CEC and the highest amount of Mo in NaHCO 3 fraction in surface layers of Kedah soil series. It has mentioned that molybdenum deficiency is not a problem in paddy fields on acid sulfate soils. [33] It can therefore be concluded that when CEC is high in surface layers of waterlogged soils, MoO −2 4 availability might be enhanced to the plants. In reducing conditions of flooded soils, molybdate is transformed to MoS −2 4 and binds to Al, Fe, and organic matter phases through sulfur bridges [34] and consequently its availability is decreased.
Because of very poor drainage status of Keranji series comparing with other series in Kedah area, increased solubility of Mo from parent material might also be accounted for higher Mo availability in such soils. It was also cited that various pastures of New Zealand, which produced forage high in Mo (>5 ppm), were found on acidic soils (pH 5.3-6.4). This could be due to the poor drainage status of soils that are intermittently wet with low redox potential during the year. [35,36] However, further studies are needed to elucidate the chemistry of Mo in poorly drained soils of paddy, and the factors influencing mobility of Mo in such soils.
The increase of CEC and surface negative charges following submergence could enhance exclusion of anions from pores, and cause reduction of impedance factor. [37] At low pH, hydroxides of iron are precipitated on kaolinite surfaces and occurrence of positive charges on theses surfaces bring about a decline in negative charges of complex. [38] In flooded conditions, sulfate reducers are active at pH values above 5, resulting in a reduction of plant available form of sulfur but at a pH less than 4.5, sulfate concentration increases in soil solution. [39] Megonigal et al., [40] illustrated that production of methane was hindered once reduction of sulfate dominated than lower ones. In Kedah area, Guar and Sedu had the highest amounts of total S at both layers. Our results are also in agreement with Paramananthan and Daud, [48] who indicated that Guar and Sedu were grouped as soils with sulfuric horizons having large levels of elemental S.

Conclusion
The results demonstrate minute concentrations of Mo fractions detected in both studied areas, but S contents were high. Only CEC and total C were significantly correlated with S and Mo contents in comparison with other soil properties. None of soil properties contributed to Mo (extractable in chemical fractions)-except CEC. Total amount of Mo had a moderate correlation with CEC in surface layers of Kedah soil series. There was no correlation between Mo and soil properties of Kelantan soil series. In contrast to Mo, we found strong and positive correlations between some soil properties (CEC and total C) and S in Kelantan soil series. No correlation was achieved between S and soil properties in Kedah area. Although there is not enough data on speciation of Mo in flooded paddy soils, it can be stated that increase of CEC in oxic soil surface layers may enhance solubility of Mo in marine alluvial soil solution. Also, the results reveal that sorption of S could be enhanced by increase in CEC and carbon contents in riverine alluvial soils.