Fine soils (<2 mm) were used for the analysis of S fractions as per the method outlined by Morche (2008). The first step in the sequential fractionation was the extraction of the soil sample with demineralized water at 1:10 (w/v). After 30 min of shaking in a rotary shaker, the samples were centrifuged at 10,000 rpm for 10 min. In the second step, the samples were extracted with 0.032 M NaH₂PO₄ at a 1:10 (w/v) ratio by shaking for 30 min followed by centrifugation at 10,000 rpm for 10 min. The residual soil sample was extracted with 1 M HCl, maintaining the extraction ratio of 1:20 (w/v). After shaking for 60 min, the samples were centrifuged for 10 min at 10,000 rpm. The extracted S fractions were measured by following the method of Chesnin and Yien (1950). The sum of all three fractions was considered the total inorganic S fraction of the soil. The total S of soil samples was analyzed using the CHNS VarioEL cube analyzer (Elementar, Germany). Organic S was calculated as the difference between the total and inorganic S of soil samples. The extractable S from soils of all land uses was extracted with six different chemical extractants (two with a neutral solution pH, two with an acidic solution pH, and two with a slightly alkaline to alkaline solution pH) having different extraction modes and chemistry (Table 2).

Schematic representation of the sequential fractionation of sulfur in soil (Morche, 2008).

Statistical analysis was performed using the Windows-based SPSS program (ver. 22.0, SPSS Inc./IBM, Armonk, United States). Duncan’s Multiple Range Test (DMRT) was used to compare means. Pearson correlation was performed between S fractions and extractable S and between soil properties to establish their relationships. A multiple linear regression equation was also computed with extractable S as the dependent variable and the main soil properties as the independent variables.

Changes in land use significantly influenced soil properties and nutrient availability (Table 3). The soil pH, measured with a 0.01 M CaCl₂ solution, ranged from 7.18 to 7.54. The highest soil pH was observed in the R-M-J system and the lowest in the R-P-G system. Soil organic carbon content (g kg⁻¹) varied between 6.68 and 7.36 across different land use systems. Compared with the R-V-J system, the R-L-S, R-M-J, and R-L-J systems showed an increase in SOC content of 6.4, 8.2, and 10.2%, respectively. Calcium carbonate content showed an increasing trend with rising soil pH, while oxides of Fe, Al, and Mn decreased with increasing pH (Table 3; Supplementary Figure S1). The iron oxide content (g kg⁻¹) ranged from 0.800 (R-M-J) to 1.063 (R-P-G). Similarly, Al_o and Mn_o contents (g kg⁻¹) ranged from 0.503 to 0.562 and from 0.342 to 0.408, respectively. Soils under all land uses had medium levels of available nitrogen content except for the R-M-J system, while they had high levels of available phosphorus and potassium. The lowest available N was found in the R-M-J system, while the highest available N was reported in the R-P-G system. However, the availability of cationic micronutrients (iron, manganese, copper, and zinc) was significantly higher in the R-L-J system compared to other land uses (Table 4).

The data presented in Table 5 show that soil organic and total S contents varied significantly among the land uses. The highest organic and total S concentrations (249.1 and 266.4 mg kg⁻¹, respectively) were observed in the R-P-G system, while the lowest (209.5 and 225.3 mg kg⁻¹, respectively) were observed in the R-L-S system. Compared to the R-L-S system, the R-V-J and R-P-G systems showed an increase in total S content to the magnitude of 11.1 and 18.2%, respectively. The organic S followed a similar pattern of occurrence as the total S content in all land uses.

A significantly higher inorganic S fraction was observed in R-P-G land use, while the lowest amount was found in the R-M-J system. The water-soluble, sorbed, and occluded S fractions together constituted the inorganic fraction. Across all land uses, the inorganic S fraction constituted 6.5% of total soil S. The water-soluble S fraction was lowest in the R-M-J system (7.1 mg kg⁻¹) and highest in the R-P-G system (11.2 mg kg⁻¹). Similarly, the sorbed S fraction ranged from 1.0 to 2.0, being lowest in the R-M-J system and highest in the R-P-G system. On the other hand, R-P-G land use witnessed the lowest amount of occluded S fraction, while R-M-J land use showed the highest value. Across the land uses, inorganic S fractions were found in the order of water-soluble>occluded>sorbed S.

Land use change significantly influenced the extractability of all six extractants (Table 6). The CaCl₂ extractable S ranged from 8.4 mg kg⁻¹ in the R-M-J system to 12.6 mg kg⁻¹ in the R-P-G system. Irrespective of the extractants used, the lowest amount of extractable S was observed in the R-M-J system, while the highest amount was observed in the R-P-G system. Interestingly, the mehlich-3 and NaHCO₃ extractants obtained on average 1.7 and 2.8 times more S than those extracted with CaCl₂ solutions, respectively. Across all land uses, the order of extractability was: NaHCO₃ > mehlich-3 > AB-DTPA > NH₄Ac-HOAc > Ca(H₂PO₄)₂ > CaCl₂.

Changes in land use significantly influence soil properties and nutrient availability. The low pH value in the R-P-G system could be due to the higher content of amorphous Fe_o and Al_o. The presence of Fe_o and Al_o generally lowers the soil pH due to their oxidation, which releases H⁺ ions into the soil system (Yang et al., 1997). Furthermore, the high quantity of chemical fertilizers applied in the R-P-G system compared to other land uses might be the reason for low soil pH. Higher SOC content in the R-M-J and R-L-J systems may be caused by the slow decomposition of organic matter due to the prevailing anaerobic environment during rice and jute cultivation. Manlay et al. (2002) also reported that anaerobic conditions could slow down the organic matter decomposition rate during rice cultivation. A high calcium carbonate content in the R-M-J system could be associated with the soil pH, which was supported by their positive correlations (S1). Similarly, the decrease of Fe_o, Al_o, and Mn_o in all land uses with the increase in soil pH can be attributed to their strong negative relationships (S1). The highest amount of available nutrient content, especially N, P, and K, may be due to the high amount of fertilizer applied in the R-P-G system compared to other land uses. The increased availability of cationic micronutrients in R-L-J land use compared to the others may correspond to the increased organic carbon content, which was observed from their significant positive correlations (S1). A reduced soil environment in rice and jute cultivations could have altered the availability of cationic micronutrients.

The highest organic and total S were found in the R-P-G system, which might be due to the lower crop removal compared to oilseed-based land use. The sulfur requirement is high for oilseed crops compared to cereals and other crops, and the inclusion of crops with high S requirement in the land use could deplete the soil S. Meena et al. (2022) reported that the inclusion of pulses in the rice-wheat land use system could influence the availability of S. Shifting the virgin soils to cultivable lands with different agricultural management could change the equilibrium status of soil organic S and affect its distribution in surface layers (Solomon et al., 2003). Organic S accounted for an average of 93.5% of the total S, representing the dominant fraction in the soils. A similar magnitude of the organic S fraction under different land uses has been reported by other researchers (Padhan et al., 2016; Meena et al., 2022). The inorganic S fraction represented the pool of S available for plant uptake. The inorganic S present in soils is readily available (extracted with distilled water or any weak salt solution), adsorbed S (extracted with phosphate-containing solution), and sulfate co-precipitated with CaCO₃ or BaCO₃ (solubilized and extracted with HCl solution). Therefore, we employed the sequential fractionation procedure as described by Morche (2008), which includes demineralized water, 0.032 M NaH₂PO_4, and 1 M HCl solution. Among the three fractions, water-soluble S accounted for an average of 58% of the total inorganic S in the soils (Figure 1). This signifies that a higher proportion of inorganic S is readily available for all land uses. On average, the magnitude of water-soluble S was 1.9 times that of occluded and 6.4 times that of sorbed S. Increased amounts of water-soluble S in the R-P-G system compared to other land uses might be due to lower crop removal.

The sorbed S varied among land uses. Among all land uses, the R-P-G system had the highest amount of sorbed S, which suggests that this particular land use had a high amount of inorganic S bound to the colloidal matrix. The presence of high amounts of Fe_o and Al_o in R-P-G could offer a specific adsorption site for SO₄⁼ (Singh, 1984; Das et al., 2020). This was confirmed by the positive relationships between the sorbed S and the oxides of Fe and Al (Supplementary Figure S2). A highly sorbed S could also suggest a high amount of bound SO₄⁼ in the soil system, as water extracts the readily available SO₄⁼ and leaves behind the adsorbed fraction. The low amount of sorbed S in the R-M-J system could be due to the organic anions released from the decomposition of soil organic matter, which could have displaced the adsorbed SO₄⁼ in the colloidal matrix of Fe_o and Al_o into the soil solution. Previous research has also shown that SO₄⁼ and low-molecular-weight organic anions compete for the same adsorption site (Evans, 1986) and that removal of organic matter can increase the adsorption capacity of soils (Das et al., 2002). Moreover, the organic anions make chelation with oxides of Fe and Al (which preferentially adsorb SO₄⁼) resulting in reduced sorption of SO₄⁼. The magnitude of occluded S was higher than that of sorbed S for all land uses. A significant positive relationship between occluded S and CaCO₃ content of soils confirmed that the presence of a high amount of carbonate could adsorb SO₄⁼, rendering it unavailable for plant nutrition (Supplementary Figure S3). It was also observed that, on average, S co-precipitated with carbonate accounted for 32.8% of the total inorganic S. Such co-precipitated S could be released and made available to the plant, which is primarily governed by rhizosphere pH. In addition, nitrogenous fertilizers applied to the soil may have a favorable effect on the solubilization of CaCO₃-occluded S, and this solubilized fraction may be partially transformed into a soluble organic form (Hu et al., 2005).

The CaCl₂ solution extracted the lowest amount compared to other extractants across land uses, accounting for 4.2% of total soil S. Suri et al. (2021) reported similar observations in soils from different land use practices. On average, Ca(H₂PO₄)₂, Mehlich-3, NH₄Ac-HOAc, AB-DTPA, and NaHCO₃ extracted 4.8, 7.3, 4.9, 12.0, and 6.2% of total soil S, respectively. The highest amount of NaHCO₃₋extractable S across land uses compared to other extractants could be due to the extraction of a portion of the organically-bound S (mostly ester sulfate) due to its high pH besides the readily available S (Kilmer and Nearpass, 1960). This was confirmed by the positive relationship between NaHCO₃-extractable S and the organic S content of the soil (computed across different land uses). In addition, the presence of HCO₃⁻ in AB-DTPA and 0.5 M NaHCO₃ displaced the sorbed S from the colloidal complex (Tabatabai, 1996), while also solubilizing partly the labile insoluble minerals in the soil (Soltanpour, 1985). Acidic extractants, particularly mehlich-3, showed higher extractability compared to neutral solvents {CaCl₂ and Ca(H₂PO₄)₂} and slightly alkaline solvents (AB-DTPA) across land uses. Low pH can solubilize the oxides of Fe and Al, thereby releasing the bound sulfate into the soil solution. Moreover, the acetate and nitrate anions in Mehlich-3 favored the extraction of S from the soil (Seth et al., 2018). The lowest amount of S extracted by a 0.15% CaCl₂ solution in soils of all land uses could be due to its efficiency in extracting only the readily available or water-soluble S. Hu et al. (2005) reported that Cl⁻-based extractants can mobilize only the water-soluble SO₄⁼ and also have lower extraction power due to their tendency to form slowly soluble CaSO₄ (Matula, 1999). Phosphate ions in Ca(H₂PO₄)₂ solution could displace the sorbed S, thereby showing increased extractability compared to CaCl₂ solution across land uses.

Among the five land uses, the R-P-G system always had the highest amount of extractable S compared to other land uses (Table 6). Moreover, the influence of the R-P-G system over the R-M-J system was more pronounced when S was extracted with 0.5 M NaHCO₃(56.0), followed by CaCl₂ (50.0), AB-DTPA (41.4), Ca(H₂PO₄)₂ (40.8), Mehlich-3 (31.2), and NH₄Ac-HOAc (30.3; Figure 2). High organic S content in the R-P-G system (Table 5), which increases the availability of S in soils (by mineralizing the labile S), was captured by the extractants used in the present study.

By establishing the relationship between extractable S (neutral extractants) and S fractions across the land uses, it was observed that CaCl₂ extractable S was positively correlated with water-soluble S (r = 0.968**), sorbed S (r = 0.950**), and organic S (0.696**), while negatively correlated with occluded S (r = −0.927**; Table 7; Figures 3A–D). A similar relationship was found between Ca(H₂PO₄)₂-extractable S and S fractions. Both extractants are highly correlated with the water-soluble S fraction, which is considered the readily available S in the majority of agricultural soils. Acidic extractants mehlich-3 and NH₄Ac-HOAc maintained significant positive correlations with water-soluble, sorbed, organic, and total S fractions while they were negatively correlated with occluded S (Table 7; Figures 4A–D). Interestingly, the NH₄Ac-HOAc-extractable S showed a high degree of correlation with organic (r = 0.914**) and total S (0.928**) compared to mehlich-3 indicating that it could extract a higher proportion from the organic S fraction. A similar relationship existed between the S extracted by weakly alkaline to alkaline extractants (NaHCO₃ and AB-DTPA) and the S fractions (Table 7; Figures 5A–D). Notably, the inorganic fraction closely corresponds to the amount of S extracted by Mehlich-3 and AB-DTPA. Occluded S was the only fraction of S that maintained a negative correlation with S obtained by the extractants across all land uses. From the relationships established between extractable S and S fractions, it can be inferred that all extractants primarily extracted the water-soluble, sorbed, and partially organic S fractions, with little extractability from the occluded S fraction.

Acidic extractants, particularly mehlich-3, were able to extract 1.7 and 1.5 times more S than CaCl₂ and Ca(H₂PO₄), respectively. Notably, the NH₄Ac-HOAc extracted higher amounts of S than CaCl₂ and Ca(H₂PO₄)₂. Across all land uses, the order of extractability was: NaHCO₃ > mehlich-3 > AB-DTPA > NH₄Ac-HoAc > Ca(H₂PO₄)₂ > CaCl₂. A Pearson correlation was constructed among the extractable S to establish their relationships. The CaCl₂-extractable S showed a strong positive correlation with Ca(H₂PO₄)₂; r = 0.988**), mehlich-3 (r = 0.967**), NH₄Ac-HoAc (r = 0.846**), NaHCO₃ (r = 0.980**), and AB-DTPA (r = 0.956**; Table 8). Such dynamic relationships among the extractable S suggested that S could be obtained from similar pools, which was further supported by their relationship with different S fractions. Comparing the procedure involved during the extraction process, it can be noted that the shaking time for mehlich-3 and AB-DTPA was restricted to 5 min and 15 min respectively, and since they are primarily used as multi-nutrient extractants, they can be employed in routine laboratory analysis. However, the unavailability of NH₄NO₃ (banned in many countries), which is one of the primary chemicals in the mehlich-3 composition, calls into question the use of mehlich-3 in routine laboratory analysis in the future. The presence of Ca in Ca(H₂PO₄) and CaCl₂ extractants flocculates the soil particles, and a clear extract can be obtained, which is highly essential for the turbidimetric determination of S. On the other hand, in the case of NaHCO₃, the high pH of the extractant could partially solubilize the organic matrix of the soil, resulting in a turbid solution that needs to be clarified prior to S determination. To test the reproducibility of extractable S by all six extractants, the relative standard deviation (RSD) value was calculated. A high RSD value indicates poor reproducibility of the data, while a low RSD value shows greater reproducibility. Across all land uses, mehlich-3 had the lowest RSD value, while NaHCO₃ showed a high RSD value signifying its poor reproducibility.

The results presented in Table 9 indicate that the extractability of different extractants is influenced by soil properties. In the case of CaCl₂-extractable S, Fe oxide alone caused a variation of 93.4%, which was improved to 95.8% with the inclusion of CaCO₃. Similarly, the variation in Ca(H₂PO₄)₂-extractable S caused by Fe oxide alone was 91.8%, and it was improved to 94.4 and 95.8% with the inclusion of Fe oxide and Mn oxide and oxides of Fe, Mn, and Al, respectively. Both CaCO₃ and Mn oxide could explain 82.8% of the variability in NH₄Ac-HOAc-extractable S, and the exclusion of Mn oxide from the regression model decreased the predictability by 8.5%. The amount of S extracted by NaHCO₃ was found to be influenced by the oxides of Fe and Al. In general, the oxides of Fe, Al, Mn, and CaCO₃ are the main soil constituents that are highly responsible for predicting the variability of extractable S in soil under different land uses.