Drying-affected soils cover about 40% of the Earth’s land surface and undergo multiple climatic stresses such as: low and seasonally variable precipitation, intense solar radiation, high evapotranspiration as well as, depending on the geographical area, relatively short summers and long winters with intense cold weather depending on the geographical area. These factors cause the occurrence of scarce biota and only allow for moderate photosynthesis Schlesinger. Although these conditions result in low soil respiration due to the low soil moisture, dryland soils account for approximately 40% of the global net primary productivity1.

Furthermore, these conditions increase the risks of desertification, salinization, land degradation, and eolian/water erosion with a strong impact on carbon (C) (including reduced sequestration), nitrogen (N), sulfur (S), phosphorus (P), and silicn (Si) cycles2,3,4,5,6,7,8. In particular, repeated drying-wetting cycles have a great impact on C, N, S, P, and Si cycling in drylands with consequent possible alteration of their conditions and vegetation decline under changing climate9,10,11,12. Oxygenic/anoxygenic photosynthesis is reported to take place in soils typically via simultaneous C and S sequestration (e.g. carbonyl sulfide, COS, uptake) by soil microorganisms13,14,15,16,17,18. However, the overall dynamics of sequestration of these biogenic components (C, N, S, P, and Si) and their subsequent mineral neoformation remain unclear in diverse soils.

On a global scale, the storage of soil organic carbon (SOC) and soil inorganic carbon (SIC) within 1-m soil depth amounts to approximately 1200–1600 and 695–940 Pg, respectively19. Interestingly, most SIC is stored in dryland soils19,20, thus these soils are effective sites for carbon sequestration via carbonate formation7,21,22,23. Furthermore, the dissolved humic substance (HS) fractions of SOC consistently serve as important reactants in the environment and are directly linked to C cycle in the global ecosystem. Drylands and particularly saline-alkaline soils are usually very rich in sulfate (SO42‒)24,25,26,27, and microbial SO42‒ reduction28,29,30 primarily controls the dissolution and mineralization of clay, Fe and layered double hydroxides (LDHs)-like minerals31,32. However, in this perspective an integrated understanding of the transformation mechanisms of organic and inorganic C, N, P, Si, S, and HS components in diverse soils is still elusive.

Microbial mineralization of soil components primarily occurs at the mineral-microbes/organic interface31,32, with labile state (LS) components representing the degradative byproducts under environmental conditions, as they are unable to the neoformation of organo-minerals, so being exported to the surrounding environments. However, some key degradative byproducts are capable of forming organo-mineral complexes that remain in the complexed state (CS) in soil33,34. But the mechanisms of generation of LS and CS soil biogenic components under environmental conditions, their possible correlation and the pivotal role played in exporting/emitting and complexing organo-minerals remains unclear.

The aim of this study was thus to clarify: (i) An integrated evaluation of biogeochemical properties for solid phase components (SOC, SIC, STN, STS, STH), labile state (LS) water-extracted components (DOCLS, HSLS, NO3‒LS, NH4 + LS, NO2‒LS, PO43‒LS, SiO32‒LS and SO42‒LS), as well as complexed state (CS) alkali-extracted components (DOCCS, HSCS, NO3‒CS, NH4 + CS, NO2‒CS, PO43‒CS, SiO32‒CS, and SO42‒CS) of four dryland soil types (forest, agriculture, grassland, and desert) in the Xinjiang Uyghur Autonomous Region that is the most important arid area in China. (ii) A comprehensive study of the biogeochemical sources, neoformation and organo-mineral state of the LS and CS HS components such as humic acids (HA), fulvic acids (FA), and protein-like substances (PLS) in diverse drying-affected soils. (iii) The assessment of SO42‒-mediated transformation mechanisms of soil components in both solid-phase and liquid-phase LS and CS forms, the sequestration of photosynthetically-derived C and S, organo-mineral neoformation and C protection, as well as corresponding exports or emissions.

Results

Biogeochemical cycling of soil organic and inorganic C with evidence from δ13C-SOC

On average, the highest SOC content was measured in forest soils (51.6 ± 42.0 g/kg) followed by grassland soils (15.6 ± 20.4 g/kg), then agricultural soils (13.8 ± 10.4 g/kg), and desert soils (1.21 ± 0.340 g/kg) (Table 1). This indicates that forest soils can store SOC arising from the repeated input of plant materials, and that soils covered by dense forests are relatively unaffected by drying effects. The SOC contribution to soil total carbon (STC), on average, followed the order: forest (66.7 ± 20.0%) > agriculture (43.7 ± 25.7%) > grassland (41.6 ± 29.0%) > desert (15.0 ± 8.4%).

Table 1 Biogeochemical properties of the four dryland soil types, i.e., forest, agriculture, grassland and desert soils, examined in this work.
Full size table

On average, the SIC content was the highest in forest soils (16.9 ± 10.0 g/kg) and it was approximately 27.0%, 42.0%, and 111.0% higher compared to, respectively, grassland, agricultural, and desert soils. This finding suggests that all soil types contribute to SIC storage and sequestration in drying-affected or arid regions. Interestingly, not only the SOC content was substantially higher in forest soils than in grassland, agricultural, and desert soils (230%, 275% and 4159%, respectively), but the SIC content was also relatively higher in forest soils than that in the other three soil types. However, the average contribution of SIC to STC followed the order: desert (85.0 ± 8.4%) > grassland (58.4 ± 29.0%) > agriculture (56.3 ± 25.7%) > forest (33.3 ± 20.0%), with the highest level in desert soil and the lowest in forest soil. Substantially higher contributions of SIC to STC in desert, grassland and agricultural soils, compared to forest soil, would correspond to the significantly lower SOC levels in the former soil types. These results imply that under either high or low SOC conditions, the accumulation of SIC in many soils via carbonate mineral formation is limited, which might control the related occurrence of metals, cations, and anions31,35,36,37,38. This further supports that desert soils act as a major store of SIC and DICLS+CS, the main fractions of which occur mostly in carbonate mineral forms7,39,40.

The stable isotopes of soil organic carbon (δ13C-SOC) values appear mostly depleted or highly enriched in forest soils (from − 26.93‰ to -21.04‰, average − 24.76 ± 2.25‰), which indicates the occurrence of C sequestration upon uptake of CO2 or DIC deriving from SOC or DOCLS+CS degradation41. On average, δ13C-SOC shows to be highly depleted in agricultural soils (-25.47 ± 0.98‰), followed by grassland (-24.85 ± 1.23‰) and forest, but highly enriched in desert soils (-23.39 ± 1.03‰). In all soil types, the δ13C-SOC values appear depleted compared to atmospheric CO2 (δ13C-DIC, 8.4‰ 42), suggesting that SOC, specifically lichen-associated cyanobacterial-fungal symbioses16,17 is not directly related to atmospheric CO2 uptake. However, the δ13C-SOC values are quite similar in the upper top layers (0–20 cm: -27.3‰ to -19.4‰) and are lower in the deeper profiles (60–120 cm: -27.5‰ to -21.7‰)23,43. Such depleted δ13C-SOC values might derive from fresh terrestrial plant materials, whilst enriched values might result from degradative carbon losses41 with subsequent generations of DIC and/or CO2 41,44,45. These results are further supported by the negative correlation between δ13C-SOC and SOC, which indicates that SOC continuously enriched in δ13C sequentially releases enriched DIC, which is subject to further sequential uptake41, resulting in a net enrichment of δ13C in SOC.

Similar to SOC, the highest contents of DOCLS and DOCCS occur in forest soils (164 ± 110 and 889 ± 572 mg/kg, respectively), which are significantly higher than, respectively, agricultural (92% and 18.0%, respectively), grassland (109% and 58.6%, respectively) and desert soils (847% and 519%, respectively; Fig. 1). The percentage contributions of soluble DOCLS and DOCCS to SOC followed the order: desert soils (1.5 ± 0.6% and 13.4 ± 10.4%, respectively) > agricultural soils (0.88 ± 0.56% and 5.6 ± 3.0%) > grassland soils (0.76 ± 0.40% and 4.5 ± 3.0%) > forest soils (0.37 ± 0.14% and 2.1 ± 1.0%), respectively. These results suggest that desert DOCLS exhibits the highest lability, indicating the highest mineral instability and degradability in both water- and alkali-extracted dissolved organic matter (DOM)LS+CS, which gradually decreases in agricultural, grassland, and forest soils. The lability of soil DOCLS (or the mineral instability) presumably results from mineral-associated SOC, which primarily derives from either plant litter46,47,48 or directly from anoxygenic photosynthetic microorganisms15. The strong correlation between DOCLS and DOCCS (Fig. 2a) indicates that the former would originate from the latter33, possibly through heterotrophic respiration at the microbe-mineral interface49,50.

Fig. 1
figure 1

Concentrations of labile state (LS) and complexed state (CS) dissolved organic carbon (DOCLS and DOCCS, a), dissolved inorganic carbon (DICLS and DICCS, b), sulfate (SO42‒LS and SO42‒CS, c), nitrate (NO3‒LS and NO3‒CS, d) in forest, agriculture, grassland and desert dry-affected soils.

Full size image
Fig. 2
figure 2

Linear Pearson correlations between some relevant parameters of dry-affected soils studied: (a) DOCLS vs. DOCCS, (b) NO3‒LS vs. NO3‒CS, (c) NH4 + LS vs. NH4 + CS, (d) DOPLS vs. DOPCS, (e) DOPCS vs. PO43‒LS+CS, (f) DOCLS vs. PO43‒CS, (g) DOCCS vs. PO43‒LS+CS, and (h) SiO32‒LS vs. SiO32‒CS. The significance of the p-values are presented at the level p < 0.01 and p < 0.001.

Full size image

Similar to DOCLS+CS, DICLS and DICCS did not reach the highest amount in the same soil type. In contrast, DICLS showed the highest level in agricultural soils (144.2 ± 67.4 mg/kg), which is significantly higher than grassland soils (68%), forest soils (86%), and desert soils (120%). Differently, DICCS showed the highest concentration in forest soils (532 ± 486 mg/kg), which is significantly higher than grassland soils (69.0%), agricultural soils (157%), and desert soils (278%). The high DICLS and the relatively low DICCS in agricultural soils would indicate that DIC in its CS form are more susceptible to mineralization in agricultural soils33, which correspondingly leads to enhanced DICLS. Furthermore, DICCS values show a significant positive correlation with DOCLS and DOCCS (Table 2), which would suggest that SOC or DOCCS mineralization33 via heterotrophic respiration49,50 could give major contribution to DIC and/or CO2 41,44,45.

Table 2 Pearson correlations between the parameters measured in 20 soil samples under four land-use types, i.e., forest, agriculture, grassland and desert, collected from the Xinjiang Uyghur autonomous region, China.
Full size table

Furthermore, the contribution of DICLS to SIC followed the order: agricultural soils (1.1 ± 0.5%) > grassland soils (0.92 ± 0.85%) > desert soils (0.86 ± 0.30%) > forest soils (0.76 ± 0.64%), while the DICCS contribution order was grassland soils (3.6 ± 5.1%) > forest soils (2.0 ± 0.9% > desert soils (2.0 ± 0.5%) > agricultural soils (1.6 ± 0.8%). These results suggest that different land uses produce distinct DICLS+CS characteristics. The fact that DICLS was much higher (~ 280%) than DOCLS in desert soils would possibly derive from the high presence of sand (45 ± 27%) rich in carbonate and silicate minerals23,51. Furthermore, sand appears to induce significant CO2 uptake via photosynthetic microorganisms, with consequent neoformation of SOC-associated minerals39,40,52,53. The significant negative correlation between DICLS and δ13C-SOC, and the insignificant correlation between DICCS and δ13C-SOC (Table 2) strongly suggests that SOC primarily arises from inorganic carbon fixation (mostly DICLS) in drying-affected soil types. Similarly, such mineral neoformation54,55 is further confirmed by the significant negative correlation between δ13C-SOC and SiO32 − CS (Table 2).

The strong correlation (p < 0.001) of DOCCS with both DICCS and DICLS, and the positive correlation of SOC with DICCS (but not with DICLS, Table S2) would suggest that DICCS mostly derives from SOC and/or DOCCS mineralization23,33. Furthermore, the strong correlation of SiO32− with both DICLS and DICCS would strongly support the at least partial origin of both DIC forms from silicate minerals, mostly clay, LDHs-like and Fe minerals-associated carbonates31,35,38. Therefore, DICLS+CS dynamics would predominantly involve, on the one side, the production resulting from dissolution and mineralization of SOC-bound clay, carbonate or LDHs-like minerals. On the other side, it would include the generation from simultaneous SOC mineralization31,35,36,56, which also leads to the export of DOMLS+CS, thereby influencing DOM dynamics in diverse soil types.

Biogeochemical cycling of STS and SO4 2- LS+CS

Soil total sulfur (STS) showed, on average, the highest levels in desert soils (8.7 ± 9.3 g/kg), which were 38.0% higher than those found in grassland soils, 444% higher than in agricultural soils, and 570% higher tahn in forest soils (Table 1). Similarly, the SO42−LS and SO42−CS fractions show significant variation among soil types (34-3.2 × 104 mg/kg and 53-5.6 × 104 mg/kg, respectively), with highest values on average in desert soils (1.4±1.1 × 104 and (1.9±1.4 × 104 mg/kg, respectively) that are respectively 103% and 47% higher than those in grassland soils, 196% and 360% higher compared to agricultural soils, and 481% and 409% higher compared to forest soils (Fig. 1c). In particular, all desert soils, one forest soil, one agricultural soil, and three grassland soils feature extremely high STS and SO42‒LS+CS values, which are key signatures of saline-alkali soils24,25,26,27.

The very high SO42−LS + CS and STS values in desert and grassland soils might arise from similar phenomena, i.e., the sulfide-dependent anoxygenic photosynthetic sequestration of S and C by COS uptake14,15,16,17,18. These processes would subsequently result in the high production and sequestration of SO42-LS+CS via redox reactions, leading to the generation of S° ↔ SO42- ↔ S2-/S22- ↔ SO42‒ (Fig. 1c)28,29,30,57,58. Significant positive correlations between SO42-LS+CS and δ13C-SOC values (Table 2; Fig. 3) suggest that sulfur sequestration occurs through SO42‒-mediated redox processes.

Fig. 3
figure 3

Pearson correlations between SO42‒LS and SO42‒CS (a), SO42‒LS+CS and the stable isotopes of δ13C-SOC (c), and SO42‒LS+CS and NO3-LS+CS (j) in drying-affected soils studied, where the significance of the p-values for SO42‒LS+CS vs. NO3‒LS+CS are at the level p < 0.001. The relationships between SO42‒LS+CS and soil relevant components in drying-affected soils studied showed that low levels of STS are associated with high levels of these components, and conversely, high levels of STS correspond to low levels of the relevant soil components. The relationships include SO42‒LS+CS with SOC (b), SIC (d), STN (e), STH (f), DOCLS+CS (g), DICLS+CS (h), SiO32-LS+CS (i), NH4 + LS+CS (k), NO2-LS+CS (l), DOPLS+CS (m), and PO43-LS+CS (n).

Full size image

The occurrence of high SO42- contents might be ascribed to several factors, such as no uptake by plants, no leaching by rainwater due to plain land, high rates of evapotranspiration, and rapid elemental conversion of S to SO42- in the presence of abundant O2 in alkaline soils58,59. In essence, high SO42−LS + CS values might have substantial impact on the SOC-associated organo-mineral complexes and on C stabilization in drying-affected soils through SOC and/or DOMLS+CS mineralization, via microbial SO42−LS + CS reduction28,29,30. This might be the key reason for extended SOC and/or DOM mineralization under drying-wetting cycles in these soil types60,61,62.

Biogeochemical cycling of STN and N-containing components

Similar to SOC, STN showed the highest content in forest soils (5.2 ± 2.7 g/kg), which was 117% higher compared to grassland soils, 190% higher compared to agricultural soils, and even 1600% higher compared to desert soils (Table 1). Correspondingly, the lowest C/N ratio was measured in forest soils (average: 15.7 ± 7.4), followed by grassland soils (19.8 ± 9.7), agricultural soils (24.3 ± 12.0), and desert soils (41.1 ± 16.0). The NO3−LS and NO3−CS concentrations varied widely (2.3-284.8 and 7.7-359.1 mg/kg, respectively) among the examined soils showing, on average, the lowest level in forest soils (57 ± 89 and 84 ± 39 mg/kg, respectively). Furthermore, the highest levels of NO3−LS occurred in desert soils (118 ± 98), followed by agricultural and grassland soils (115 ± 95 and 114 ± 136 mg/kg, respectively), while the highest NO3−CS occurred in agricultural soils (142 ± 151 mg/kg), followed by grassland and desert soils (106 ± 118 and 97 ± 87 mg/kg, respectively (Fig. 1d). The high variation of NO3−LS + CS contents in the different soil types might arise from drying-rewetting N-mineralization processes, whereas NO3− can remain stable and steady for more than a month after the last stress by rewetting-drying63. Similarly, drying-wetting cycles may cause a decrease of mineral N and an increase or decrease of NO3-, depending on soil types, C compounds, and drying intensity63,64,65, which are possibly related to a decline of soil fungi under dry conditions63,64. In particular, the low levels of NO3‒LS+CS in forest soils might arise from high N mineralization and gaseous losses of N via nitrification followed by denitrification62,63,64, possibly caused by the availability of water inside soils covered by dense forest.

A significant positive correlation between NO3‒LS and NO3‒CS, and between each of them and STS and SO42-LS+CS (Fig. 1b; Table 2), would suggest that NO3‒LS and NO3‒CS originate from similar sources by a two-step pathway: there is NH4+ production25,66 from SOC or DOM respiration and mineralization via microbial SO42-LS+CS reduction28,29,30, followed by autotrophic NH4+ oxidation into NO3‒ by chemolithoautotrophs67,68. The NH4 + LS and NH4 + CS contents vary largely among soil types (2.2–22.2 and 5.1–52.3 mg/kg, respectively), being highest on average in forest soils (10.0 ± 7.8 and 30.7 ± 20.8 mg/kg, respectively) and lowest in desert soils (3.3 ± 0.6 and 5.6 ± 0.4 mg/kg, respectively; Fig. 4b). The high level of NH4 + LS+CS in forest soils might be due to the dissolution and mineralization of NH4+-bound minerals and N-bound SOC and/or DOM31,38,69,70, which remains stable due to the relatively low presence of SO42‒ in forest soils (Fig. 1c). Furthermore, clay contents show a significant positive correlation with NH4 + CS (Table 2), which suggests that clay minerals might be a source of NH4[+71. The strong positive correlation between NH4 + LS+ and NH4 + CS (Fig. 2c) suggests that both would likely originate from similar sources. Interestingly, the low NH4 + LS+CS in desert soils might be related to the abundance of SO42− and NO3‒LS+CS (Fig. 1d). These results are further supported by the significant negative correlation of SO42−LS with NH4 + CS and by the significant positive correlation between SO42−CS and NO3‒LS+CS (Table 2). Therefore, SO42- plays pivotal roles in the steady-state occurrence and levels of NO3‒LS+CS and NH4 + LS+CS. The NO2 − LS and NO2 − CS concentrations appeared relatively low in all soil types (0.2-5.0 and 0.1–6.8 mg/kg, respectively) and were, on average, the lowest in desert soils (0.21 ± 0.03 and 0.17 ± 0.03 mg/kg, respectively), and the highest in agricultural soils (NO2 − LS, 2.8 ± 1.9 mg/kg) and forest soils (NO2 − CS, 3.2 ± 2.6 mg/kg) (Fig. 4a). The remaining text of this subsection is discussed in the SI.

Fig. 4
figure 4

Concentrations of labile state (LS) and complexed state (CS) nitrite (NO2‒LS and NO2‒CS, a), ammonium (NH4 + LS and NH4 + CS, b), phosphate (PO43‒LS and PO43‒CS, c), dissolved organic phosphorus (DOPLS and DOPCS, d) and silicate ion (SiO32‒LS and SiO32‒CS, e) in forest, agriculture, grassland and desert dry-affected soils.

Full size image

Biogeochemical cycling of dissolved organic phosphorus and PO4 3 − LS+CS

The PO43−LS and PO43−CS concentrations showed wide variation (0.5–17.5 and 0.3–13.0 mg/kg, respectively), with the highest average PO43−LS level in agricultural soils (9.4 ± 8.1 mg/kg), which is 91% higher compared to grassland soils, 168% higher compared to forest soils, and 1220% higher compared to desert soils (Fig. 4c). Differently, PO43−CS was the highest (7.2 ± 5.1 mg/kg) in forest soils, that is 94% higher compared to grassland soils, 130% higher compared to agricultural soils, and 1200% higher compared to desert soils. The dissolved organic phosphorus (DOP)LS value (0.3–18.1 mg/kg) was a bit lower than DOPCS (0.8–22.2 mg/kg), with DOPLS showing the highest value (8.1 ± 8.7 mg/kg) in agricultural soils, namely 310% higher than in forest soils and 340% higher than in grassland soils, while DOPLS was undetectable in desert soils (Fig. 4d). Differently, DOPCS was highest (14.4 ± 7.3 mg/kg) in forest soils, which is 49% higher than in agricultural soils, 86% higher than in grassland soils, and 1250% higher than in desert soils. In particular, high PO43−LS in agricultural soil might arise from manure application, whereas high PO43−CS in forest soil might arise from two possible sources via redox cycling57: (i) dissolution and mineralization of phosphate minerals35,57,72,73,74, and (ii) mineralization of DOP–bound SOC- or DOC-associated minerals57,73,74,75. The latter process is further supported by the significant correlations of DOPCS with PO43 − LS and PO43 − CS (Fig. 2e), and of DOCLS (as well as DOCCS) with PO43 − LS and PO43 − CS (Fig. 2f-g; Table 2). Remarkably, low STS and SO42‒LS+CS contents were well correlated with high amounts of DOPLS+CS, and PO43‒LS+CS, and vice versa (Figs. 3m, n and 5j, k). These results would suggest that STS might be involved in DOP mineralization57,73,74,75 through SO42 − LS+CS-mediated reduction processes28,31, whereas the produced PO43‒ might be associated to the neoformation of clay and secondary minerals31,35,38.

Fig. 5
figure 5

Relationships between soil total sulfur (STS) and soil relevant components in drying-affected soils studied, showing that low levels of STS are associated with high levels of these components, and conversely, high levels of STS correspond to low levels of the relevant soil components. The relationships shown include STS with SOC (a), SIC (b), STN (c), STH (d), DOCLS+CS (e), DICLS+CS (f), SiO32-LS+CS (g), NH4 + LS+CS (h), NO2-LS+CS (i), PO43-LS+CS (j), and DOPLS+CS (k). Pearson correlation of STS with NO3-LS+CS (l) in drying-affected soils studied. The significance of the p-values are at the level p < 0.01 or p < 0.001.

Full size image

Biogeochemical cycling of SiO3 2- LS+CS

The SiO32-LS and SiO32-CS contents differed substantially among the studied soils ranging, respectively, between 57 and 328 mg/kg and between 13 and 924 mg/kg, with the highest average values in agricultural soils (210 ± 109 and 565 ± 404 mg/kg, respectively). These values were higher by, respectively, 61% and 65% compared to forest soils, 60% and 89% compared to grassland soils, and 87% and 416% compared to desert soils (Fig. 4e). The remaining text of this subsection is discussed in the SI.

Transformation features of humic substances (HS) components

The EEM-PARAFAC model could identify one to three fluorescent components in the water-extracted and alkali-extracted DOMCS or HSCS of the four studied soil types (Fig. 6; Table S2). The three components can be denoted as humic acids (HA), fulvic acids (FA) and protein-like substances (PLS), which collectively define the HS composition76,77. Forest soils are characterized by only one component denoted as degraded HALS with two peaks in a relatively short wavelength region (Table S2). This implies that the other two components, i.e., FALS and PLSLS of HSLS are completely degraded or mineralized (Fig. 6), thereby suggesting the highly labile and degradative nature of forest HSLS components. Differently, three components (HACS, FACS and PLSCS) of HSCS are present in forest soils, with two long-wavelength peaks for HACS and FACS, and two protein-like adjacent peaks (260/299, 340 nm) (Fig. 6). These results suggest that HSCS components remain mostly unaffected, i.e., their recalcitrant nature in forest soil, presumably their involvement in organo-minerals complexes37,78,79.

Fig. 6
figure 6

Fluorescence (excitation-emission matrix, EEM) spectra of water-extracted/labile state (HSLS) (a) and alkali-extracted/complexed state (HSCS) (b) of humic substances from forest, agriculture, grassland and desert dry-affected soils, and detected fluorescence peaks of humic acid-like (HA-like), fulvic acid-like (FA-like) and protein-like substances-like (PLS-like) identified using EEM-PARAFAC modeling. ‘nd’: not detected.

Full size image

Agricultural soils featured two components, i.e., degraded HALS with two peaks and PLSLS with two longer-wavelength peaks (M and T; Table S2 and Fig. 6). These results suggest that HSLS or DOM in agricultural soils may have been degraded due to drying effects. By contrast, HACS and FACS in agricultural soils show the peaks (C, M, and A) in relatively long-wavelength regions, implying that these two components remain primarily unaffected as in the case of forest soils. This finding might suggest that HA- and FA-bound organo-minerals in CS forms remain unaffected under extreme drying conditions, as reported in earlier studies31,79,80. Differently, the PLSCS showed the complete disappearance of the four peaks of humic-like and protein-like fractions, which subsequently lead to the production of tryptophan-like substances with two new peaks (T, 275/351 nm and peak TUV, 220/351 nm81,82. These results suggest that tryptophan-like substances might be secondary derivatives originating from degradation of PLS in agricultural soils by microbial activities83,84. The remaining text of this section are presented in the SI.

Discussion

Photosynthetically-derived SOC and pH-dependent SIC or DICLS+CS sequestration

The significant negative correlation between DICLS and δ13C-SOC, and the insignificant correlation between DICCS and δ13C-SOC (Table 2) strongly suggests that SOC primarily arises from inorganic carbon fixation (mostly DICLS) in drying-affected soil types. Such photosynthetically-derived SOC exports HS components that are mainly responsible for mineral neoformation54,55, which is further confirmed by the significant negative correlation between δ13C-SOC and SiO32-CS (Table 2). Further evidence of such photosynthetically-derived SOC is provided by the high variation of δ13C-SOC (-26.93‰ to -21.04‰) among all soils (Table 1), which might derive from uptake of DIC originating from the heterotrophic mineralization of HS components46,48,85. With this regard, many studies have found that anoxygenic photosynthesis generally occurs via simultaneous carbon and sulfur sequestration by lichen-associated cyanobacterial-fungal symbioses, resulting in COS uptake[13–18. Further evidence of SOC-associated sulfur sequestration is provided by the significant correlation of δ13C-SOC with both SO42-CS and SO42-LS (Table 2).

Most importantly, because of the absence of any plant material in desert soils, newly-derived FALS and PLSLS (Fig. 6; Table S2) would primarily arise from SOC produced by anoxygenic photosynthesis and/or primary producers39,40,52,53. Such anoxygenic photosynthetic activities are further supported by the detection of newly-derived, intact and fresh PLSLS+CS in desert soils (Fig. 6), similar to those found in forest soils77. Further support to this interpretation is provided by the occurrence of mineral-bound degraded FACS and PLSCS as discussed above (Fig. 6; Table S2)77. Moreover, desert’s FALS and PLSLS components are absent in either forest or agricultural soils, where plant material-derived FALS and PLSLS do not form rapidly. In contrast, anoxygenic photosynthetic microorganisms may rapidly produce fresh FALS and PLSLS as in the case of water and snow environments82,86. However, FACS and PLSCS derived from both plants and anoxygenic photosynthesis would remain in HSCS-bound minerals36,37,79,87, which implies the stabilization of SOC or HSCS via organo-mineral formation in extreme drying-affected soils, except for grassland soils that appear to be highly vulnerable for C stabilization and accumulation via organo-mineral complexes. Therefore, anoxygenic photosynthesis in desert soils might contribute DOM or HS through soil microorganisms and generate carbonate minerals via sequestration and uptake of DICLS+CS and COS derived from SOC and DOC mineralization and/or atmospheric COS. These processes would play important roles in reducing CO2 emissions from soil.

The significant negative correlations of pH with SIC and DICCS, the significant positive correlations of SIC with SOC, STN, STH, and water contents, and of DICCS with water contents and SOC (Table 2; Fig. S4a) would suggest a pH-dependent SIC or DIC uptake in the sequestration and fixation of photosynthetically-derived SOC under wet conditions16,32,88,89. Furthermore, the significant negative correlation of pH with SOC and DOCLS+CS (Fig. S4b; Table S2) would suggest that high SOC up to a certain acidic pH contributes primarily to all microbial activities due to the relatively low contribution of DOCLS+CS (approximately 1.4–33.6% and, on average, 7.5 ± 7.3% of SOC). High microbial activities at near-neutral pH (5.5–7.5) might be a key factor for the high occurrence of SOC, ascribed to high bacterial diversity and biomass90,91, due to the high availability of nutrients and HS present in weak (when formed) organo-mineral complexes with metals90,92,93. Remarkably, a mild acidic pH would imply soil HS to remain in a non-complexed state, which is further supported by the significant negative correlations of pH with SOC, STN, STH, DOCLS+CS, DICCS, NH4 + CS, NO2‒CS, PO43‒CS and DOPCS (Table 2). These results also suggest higher availability of all these components up to acidic pH 5.63 83,94,95, which might be due to intense fungal and bacterial activities90,91. Mild acidic-alkaline pH values would lead to high SOC sequestration by lichen-associated cyanobacterial-fungal symbioses16,17,18,32,88,89, due to the higher presence of various nutrients plus SIC or DIC96. This, in turn, would imply a high heterotrophic SOC mineralization due to the higher presence of bacteria under near-neutral pHs49,83,97 via SO42 − LS+CS reduction during simultaneous sulfur and carbon sequestration. Such simultaneous sequestration-mineralization processes under near-neutral pH conditions would thus lead to the presence of high SIC, SOC and STN contents, together with their dissolved forms such as DICCS, DOCLS+CS, NH4 + CS, NO2‒CS, PO43‒CS, and DOPCS (Fig. 2; Table 1).

Differently, an increase of pH (> pH 7.5) would primarily result in the formation of strong organo-mineral complexes, which would occur due to the increased presence of anionic forms (e.g. –COO–) of the functional groups in HS components under elevated pH conditions36,98. In turn, this process would induce a reduced availability of SIC and most mineral nutrients96. These conditions would involve the simultaneous decrease of the diversity of soil bacterial communities and enzymatic activities (Table 2)99. Thus, the main effects would include a decrease in sulfate-reducing bacteria100 due to reduced nutrient availability, which results from increased SOC sequestration and stability under high pH conditions16,18,88. Soil C sequestration via mineral neoformation is further supported by the significant negative correlation of δ13C-SOC with DICLS, NH4 + LS, NO2‒LS+CS, SiO32‒CS, and PO43‒LS, possibly due to their anoxygenic photosynthetic uptake13,15,16,17. Thus, pH plays an important role in regulating SIC and SOC sequestration-transformation via DICLS+CS, which is primarily controlled by the occurrence of DOCLS+CS and SiO32 − LS+CS in the various soil types examined.

Remarkably, the SiO32−CS-normalized DICLS and DICCS values, i.e., the DICLS/SiO32−CS and DICCS/SiO32−CS ratios showed a significant linear relationship between them (Fig. S4g) whereas the DICCS/SiO32−CS ratio varied largely, from 0.2 to 6.0 mg/kg/SiO32−CS and showed, on average, high and very similar values in forest soils (2.0 ± 2.0 mg/kg/SiO32-CS), grassland soils (1.9 ± 2.1 mg/kg/SiO32−CS), and desert soils (2.1 ± 1.37 mg/kg/SiO32-CS), while the lowest value was found in agricultural soils (0.6 ± 0.5 mg/kg/SiO32−CS). These results would suggest that DICCS in forest, grassland, and desert soils remains primarily under mineral protection, mostly in carbonate and secondary clay, Fe and LDHs-like minerals. Interestingly, the lowest DICCS/SiO32−CS ratios and the highest DICLS/SiO32−LS ratios measured in agricultural soils would indicate that the dissolution and mineralization of silicate minerals19,101,102 might intensively occur during human activities (e.g. ploughing, tilling, and cultivation), thereby resulting the mineralization of DICCS from minerals and its subsequent export into DICLS only in agricultural soils85. These results are further supported by the significant negative correlation between pH and DICCS/SiO32-LS ratios (Fig. S4j).

Mechanisms of SO4 2−-mediated mineralization of soil biogenic components

The SO42−LS and SO42−CS concentrations follow the order: desert soils (respectively, 103.2% and 47.4% higher than grassland soils) > grassland soils (respectively, 45.4% and 212.4% higher than agricultural soils) > agricultural soils (respectively, 96.5% and 10.6% higher than forest soils) > forest soils. In essence, a comprehensive relationship of STS and SO42‒LS+CS with soil biogenic components occur in either solid phase (e.g., SOC, δ13C-SOC, SIC, STN and STH) or liquid phase (e.g. DOCLS+CS, DICLS+CS, SiO32‒LS+CS, NH4 + LS+CS, PO43‒LS+CS and DOPLS+CS), showing low STS and SO42‒LS+CS contents correlate with high amounts of soil components in the solid and liquid phases for all soil type, and vice versa (Figs. 3 and 5). These results imply that intense redox activities would arise from sulfur and carbon sequestration by COS and/or H2S uptake28,57,103 through sequestration (S2− ↔ S°) and reduction-oxidation (redox: S° ↔ SO42− ↔ S2−/S22− ↔ SO42‒)28,29,30 by lichen-associated cyanobacterial-fungal symbioses14,15,16,17,28,104. Specifically, significant positive correlations of SO42‒LS and SO42‒CS with δ13C-SOC (Table 2; Fig. 3c) indicates the occurrence of aforementioned C and S sequestration which simultaneously maintains STS via SO42‒ redox processes.

Transformations mediated by SO42− are further supported by the following results. First, Principal Component Analysis (PCA) of the solid- and liquid-phase components has revealed the existence of significant consistent associations of negative loadings of STS, SO42‒LS SO42‒CS, δ12C-SOC, NO3‒LS, NO3‒CS and pH with positive loadings of SOC, SIC, STN, STH, DOCLS, DOCCS, DICLS, DICCS, NH4 + LS, NH4 + CS and water content for PC1, which accounts for the highest variability (47.3%; Table 3). These results suggested that SO42‒ reduction was very relevant in enhancing the sequestration of both inorganic and organic C and N in the growth of photosynthetic soil microorganisms under conditions of high water availability and low pH. Concurrently this process would also facilitate the release of DOC, DIC, and NH4+. The observed increased sequestration of C and N was further supported by the analysis of PC2, which indicated that the positive loadings of SOC, δ12C-SOC, STN, STH, STS, SO42‒LS and SO42‒CS were associated to the uptake of DICLS, SiO32‒LS and SiO32‒CS, as indicated by their negative loadings (Table 3). Furthermore, the occurrence of C sequestration, along with the stability of organo-mineral and carbonate mineral, was supported by the significant negative loadings of δ12C-SOC associated to the significant positive loadings of DOCCS and DICCS, as indicated by PC3 and PC5 (Table 1). In conclusion, the results described above indicated the simultaneous occurrence of sequestration and SO42‒CS reduction, which are consistent with results of previous studies16,28,89,96,103,105,106.

Table 3 Loadings of soil component parameters on five significant principal component variables and total variance explained, for samples collected from four drying-affected soil types (forest, agriculture, grassland, and desert) in the Xinjiang Uyghur autonomous region, China.
Full size table

Second, results from an experimental study conducted on a forest haplic luvisol have demonstrated that the sequestration of C, N, and S by soil microorganisms might occur simultaneously during their growth and metabolism, and result in a net increase of STS (1.2–41.0%) and STN (1.7–7.0%) under sunlight, dark, and control conditions during the incubation periods of 0, 30, 75, and 150 days34. However, C sequestration was not consistently detected due to the simultaneous offsets of photochemical and microbial respiration, which led to a significant loss of SOC. In particular, a net increase of SOC was measured only under sunlight (0.80%) and control (0.40%) conditions during the 75–150 day period following mineralization, whereas a decrease from 1.8 to 2.4% and from 2.3 to 6.0% respectively, was measured during the initial 0–75 days34.

Finally, a similar sequestration of C, N and S, together with SO42−-mediated transformations, have been measured in various soils and desert sands of Inner Mongolia, for which significant positive correlations of STS with SOC and STN, and significant negative correlations with δ13C-SOC (ranging from − 24.9‰ to -7.10‰) were observed96.

Such photosynthetic S and C fixation by microorganisms is substantially stimulated in wet conditions104,105,106 and is the only primary way by which they conserve their energy, diversity, and biomass from sulfide oxidation in desert soils104,107. In parallel, the eight electrons released from the redox reactions (S2–] → SO + 2e‒ and SO → SO42‒ + 6e‒)30 would generate reactive oxygen species and trigger export of soil components from the mineralization/dissolution of various minerals83,90,108,109. Moreover, the linear positive relationships of both NO3‒LS and NO3‒CS with STS, SO42‒LS, and SO42‒CS would suggest a simultaneous occurrence of both nitrification and denitrification62,63,64. Furthermore, the DOCLS+CS mineralization33 determines the export of several soil components (e.g. DICLS+CS, NH4 + LS+CS, SiO3LS + CS, etc.) by mineral dissolution31,35,56,69, as well as of DICLS+CS and NH4 + LS+CS from DOCLS+CS mineralization25,66,82 via SO42 − LS+CS reduction. In particular, the SO42−LS-normalized DOCLS and DOCCS show relatively higher contents (0.3–3100 and 1.1–14800 g/kg/SO42−LS, respectively) compared to the corresponding SO42−CS-normalized DOCLS and DOCCS values (0.16–570 and 0.52–5900 g/kg/SO42−CS, respectively). Very similar results are obtained for DICLS and DICCS (0.6–1900 and 1.4–9400 g/kg/SO42−LS), which are respectively higher than 0.3–1000 and 0.7–3000 g/kg/SO42−CS, and for SiO32−LS and SiO32−CS (1.6–1700 and 0.2–5100 g/kg/SO42−LS, respectively, higher than 0.8–1500 and 0.1–4900 g/kg/SO42−CS), as well as for other components (Fig. 3). These effects are presumably due to the occurrence of mineralization and respiration at the mineral-microbes/organic interface31,32 or of CS forms in soils33, which subsequently cause the export of soil components in LS-forms (Fig. 7). These results would indicate that SO42‒LS+CS-mediated redox processes28,29,57 primarily lead to the dissolution of HS-associated clay, Fe, and LDHs-like minerals, so that the existing organo-mineral soil components would remain in a degraded state under the existing environmental factors and conditions77,110 after undergoing mineralization by SO42‒LS+CS-mediated microbial heterotrophic respiration28,29,49,57. The soil SO42−LS + CS contents ultimately lead to control the mineral-associated soil components as depicted in the conceptual model shown in Fig. 7.

Fig. 7
figure 7

A conceptual model for the sequestration of C, S, and N, which highlights the simultaneous maintenance of sequestration (S2- ↔ S°) and reduction-oxidation (redox: S° ↔ SO42- ↔ S2-/S22- ↔ SO42‒) processes by soil microorganisms during the transformation of organo-mineral components in the drying-affected soils. This leads to the exports of their LS-forms (DOCLS, HSLS, NO3‒LS, NH4 + LS, NO2‒LS, PO43‒LS, SiO32‒LS and SO42‒LS) and mineral neoformation of their CS-forms (DOCCS/HSCS, NO3‒CS, NH4 + CS, NO2‒CS, PO43‒CS, SiO32‒CS and SO42‒CS).

Full size image

Apparently, the SOC production by anoxygenic photosynthesis in desert soils39,40,53 and by decay of from plant-derived litter materials47,48 and subsequent mineralization by SO42 − LS+CS and/or STS could be a common unique feature of all soil types (Figs. 3 and 5). The transformation (e.g., mineralization or neoformation) of soil components via SO42‒ reduction leads to the subsequent export of components in LS, CS, and mineral forms (Fig. 7) to ambient environments through rainwater and water discharge. These processes would suggest that SO42‒LS+CS–mediated redox reactions are the key controlling factors for the dissolution of clay, Fe and LDHs-like minerals, and for maintaining all soil components in steady state with SO42 − LS+CS and STS.

Signals of soil desertification and salinization

The degradative steady state of SOC, SIC, STN, and STH in the various examined soils, due to STS-mediated transformation, is primarily responsible for the coherent transformation of SOM77,110. The occurrence of this degradative state is further supported by the corresponding dissolved-phase components (which include DOCLS+CS, DICLS+CS, SiO32‒LS+CS, NH4 + LS+CS, NO2−LS + CS, PO43−LS + CS, and DOPLS+CS), and by the SO42‒LS+CS-mediated heterotrophic respiration in the soil matrix28,29,49,57. Such solid-phase and liquid-phase transformations clearly suggest that STS via SO42‒LS+CS reduction-oxidation (redox) processes28,29,57 are key factors in regulating the relevant soil components. Such transformations would induce degradation of mineral-associated SOC and DOCLS+CS, and simultaneously produce DICLS+CS and other nutrients. The continuous degradation of SOC without input from decaying plant materials46,47,48 leads to increasing SIC via inorganic carbonate mineral neoformation21,22,23,51. This process unveils the highest contents of STS and SO42‒LS+CS at low contents of SOC and nutrients, which might initiate desertification.

Remarkably, the repeated occurrence of sulfide-dependent anoxygenic and oxygenic photosynthetic sulfur and carbon sequestration through COS uptake18 by lichen-associated cyanobacterial-fungal symbioses14,15,16,17,104 causes accumulation of S° with the simultaneous transfer to SO42- by soil fungi and bacteria28. In particular, desert soils feature a significant CO2/DIC uptake via photosynthetic microorganisms and, possibly, SOC-associated mineral formation39,40,52,53. Notably, high SO42‒ contents are often detected in saline-alkali soils that are usually denoted as sulfate-type saline soil24,26,27. Therefore, high contents of STS and/or SO42‒LS+CS in one forest soil and one agricultural soil, three grassland soils, and all desert soils would represent the prime signature of increasing salinity and desertification.

Materials and methods

Soil sampling sites, collection, and Preparation

The studied soils were collected from the Xinjiang Uygur Autonomous Region located in the northwest part of China, i.e., the central part of Eurasia (73°40′–96°23′E longitude and 34°25′–49°10′N latitude), which covers an area of more than 1.66 million km2 that makes Xinjiang the largest province in China. The mountainous area includes hills and plateaus for approximately 800,000 km2, and the plain area includes the intermountain Tarim and Junggar Basins for approximately 800,000 km2. The climate of Xinjiang is temperate continental with a long winter (mid-Oct-Mar) and a relatively short summer (May–Aug). The air temperature is largely fluctuating showing the highest value of 41.0 °C and the lowest one of − 28 °C (Table S1). Xinjiang is an arid area featuring small and uneven annual precipitations that ranged from 74.0 to 409 mm since 2012, showing the lowest value in 2022 and the highest in 2015 (Table S1).

The land use/cover of Xinjiang includes six categories: cultivated and arable land (5.5%); forest land (1.68%); grassland (29.5%); water area (2.1%), construction land (0.6%), and unused land (60.5%)114. More than half of soils in Xinjiang consist of sandy, desert (Gobi), saline-alkaline, bare soils and bare rocks. The mountainous area is more extended than the plain one, and the northern area is more extended than the southern one. The dominant forest tree species is populus Euphratica, and forest types include valley secondary forest and plain artificial forest. The ecological green component of forest land, grassland and cultivated land accounts for just 37.2% of the total land area of Xinjiang. This indicates that its ecological and environmental quality is not high, the system resistance is not strong, and the ecosystem is fragile.

In this work a total of 20 soil samples, i.e., 5 under forest, 4 from agricultural land, 6 from grassland, and 5 from desert were collected near Urumqi City, (Fig. S1), in a sampling area ranging from N 42°31′04″–N 45°27′36″, E 86°17′34″4–E 88°20′51″.

The forest soils were collected under coniferous species located near the slum areas of Urumqi. The grassland soils were mostly collected from artificial grassland and scattered in Urumqi and its surrounding areas. The desert soils were characterized by the absence of vegetation cover and high sand content, and some are red soils in sandy areas. The agricultural soils were collected from soybean, rape, and corn fields.

Samples were randomly collected from the soil top layer (~ 0–15 cm) after removing grasses and debris, using a 10-cm diameter stainless steel auger as described previously77. After removing manually plant litters, roots, and debris, each sample was mixed homogeneously to obtain a spatially representative sample at the field scale. Soil samples were then oven dried at 40 °C77, passed through a 2-mm sieve, ground in a mortar to obtain particle sizes lower than 0.2 mm, and stored at -20 °C until further use.

Water and alkaline extracts of soil samples

Soil HS were isolated using a novel method77 based on the extraction with water first, followed by a 0.1 M NaOH solution. The two extracts were obtained by several subsequent steps (Fig. S1). In the first step, ultrapure water (18.2 MΩ·cm, Mill-Q, Millipore) was added to each ground and sieved soil sample at a 1:10 soil/water ratio, the mixture was vortexed for 1 min in closed 500-mL brown bottles, and then shaken for 6 h at 25 °C. Afterwards, it was centrifuged for 20 min at 4000 rpm using a Thermo Fisher Scientific SORVALL ST 16 centrifuge, to remove suspended solids. The supernatant solution was then filtered through a 0.45-µm membrane filter (GF/F type, Shanghai Xin Ya Purification Equipment Co. Ltd, China), whereas the remaining solid residue was extracted again with fresh ultrapure water for 1 h. The above procedure was applied again to obtain a supernatant solution that was mixed with the previous one and stored in freezer at − 20 °C until further processing. This solution represents the soil water extract (We).

In the second step, the soil residue after water extraction was subjected to alkaline extraction under N2 with a 0.1 M NaOH solution, at 1:10 soil residue/alkaline solution ratio, by shaking for 3 h at 25 °C. The mixture was then centrifuged as described above and the supernatant solution filtered through a 0.45-µm membrane filter (polytetrafluoroethylene membrane, PTFE, Shanghai Xin Ya Purification Equipment Co. Ltd, China). The remaining solid residue was extracted again with a fresh alkaline solution for 3 h, and the procedure described above was applied again to obtain another supernatant solution that was mixed with the previous one to obtain the alkaline extract (Ae), and stored in freezer at − 20 °C until further processing.

The We and Ae samples represent, respectively, the water-extracted (labile state, LS) components (DOM/DOCLS, HSLS, NO3‒LS, NH4 + LS, NO2‒LS, PO43‒LS, SiO32‒LS and SO42‒LS) and the alkali-extracted (complexed state, CS) components (correspondingly, DOCCS, HSCS, NO3‒CS, NH4 + CS, NO2‒CS, PO43‒CS, SiO32‒CS, and SO42‒CS).

All glass bottles used for extraction purposes were soaked with 10% HNO3 and HCl for 24 h before use, rinsed with Mill-Q (MQ) water, and then placed in a muffle furnace for 4 h at 550 °C in order to remove residual organic matter. Similarly, each glass fiber filter was individually wrapped in an aluminum foil and then placed in a muffle furnace for 4 h at 550 °C, in order to remove any organic and inorganic impurities.

Analytical methods

Soil particle size was measured on the 2-mm size soil fraction using the hydrometer method (Mastersizer 3000, Malvern). Soil water extracts were directly used for all analytical measurements (DOC, nutrients, fluorescence), whilst alkaline extracts were previously adjusted to approximately pH 8.0 before measurements. The DOC and DIC contents in We and Ae samples were measured in triplicate by injection of each sample into an Aurora combustion total organic carbon (TOC) analyzer equipped with autosampler (O.I. Analytical Aurora-Model 1030 W, USA). The pH and EC were measured in a 1:2.5 soil residue/ultrapure water solution ratio, using a pH meter (Multi 3630 IDS, Germany) equipped with a combination electrode.

Nutrients, including total phosphorus (TP), phosphates (PO43-), total nitrogen (TN), nitrates (NO3-), nitrites (NO2-), ammonium (NH4+), and silicates (SiO32-), were determined in We and Ae using an automated continuous flow analyzer system (SKALAR San++, Netherlands).

Soil total carbon (STC), soil total nitrogen (STN), soil total sulfur (STS) and soil total hydrogen (STH) were measured with a Vario EL Cube Elemental Analyzer (Elementar VarioEL III, Germany) on approximately 20-mg aliquots of dried, ground and homogenized sample contained into a clean, carbon-free pre-combusted tin boat placed in the autosampler rack assembly. Soil organic carbon (SOC) was analyzed by an Elemental Analyzer described elsewhere76 on dried, ground, and homogenized samples pretreated with 1 M HCl, which were contained in vials in the autosampler tray. Sulfanilamide was used as standard after every 10 measurements. Soil inorganic carbon (SIC) was obtained as SIC = STC - SOC. The stable 13C isotope was analyzed by an Elemental Analyzer (Flash 2000 HT) interfaced with a stable carbon isotope ratio mass spectrophotometer (MAT 253 PlusTM, ThermoFisher, USA) on each soil sample, prepared as described above and placed in a IAEA-600 autosampler rack assembly, using caffeine as the 13C standard.

The remaining text of site description, water and alkaline extracts of soil samples, analytical methods (fluorescence spectra and PARAFAC model), and statistical analyses are discussed in the supplementary information.