Abstract
Introduction:
Dryland soils are highly sensitive to land use change and restoration, yet the relative influence of land use and associated vegetation and management conditions on soil dissolved organic matter (DOM) remains poorly understood.
Methods:
This study evaluated DOM quantity and quality across five land use types (cropland, orchard, grassland, shrubland and forestland) and the additional differences associated with vegetation and management within specific land use systems. Soils were sampled from sixteen sites and three depths (0-20, 30-50, and 60-80 cm), including naturally restored and artificially planted shrublands, grasslands with taproot and fibrous-root systems, orchards under high-input and low-input management.
Results:
Dissolved organic carbon (DOC) and DOM optical properties were characterized using ultraviolet-visible absorbance and fluorescence spectroscopy. Overall, DOC content was highest in forestland topsoil (235.22 mg kg−1) and lowest in orchards and cropland (10.16-89.15 mg kg−1), and declined with depth. Humic-like components dominated the fluorescence signal, whereas protein-like components increased in deeper layers. Woody systems showed higher optical signatures consistent with larger apparent molecular size, greater aromaticity, and stronger humification, whereas cropland showed the opposite pattern. Within land use types, the walnut/low-input orchard showed higher DOC and stronger humic-like optical signals than the apple/high-input orchard, whereas grassland root type and shrubland restoration mode showed weaker changes. Multivariate analysis showed that DOC content and DOM composition covaried with surface inputs, fine texture, and biological soil crust development.
Discussion:
This framework provides a useful basis for interpreting soil carbon responses to restoration and land management in fragile dryland ecosystems.
1 Introduction
Drylands occupy approximately 41% of the global land surface, and their ecosystem functioning is highly vulnerable to climate change and human disturbance (Prăvălie, 2016). Under such sensitivity, relatively small changes in soil carbon pools can induce shifts in ecosystem services and biogeochemical regulation at regional to even global scales (; Schimel, 2010). Land use change and management practices often interact in dryland landscapes, making it necessary to consider their combined influence when evaluating soil carbon cycling (; Prăvălie, 2016). Soil organic carbon (SOC) is widely adopted as a primary indicator of soil quality, fertility maintenance, and carbon sequestration potential (). Yet SOC responses to land use and management are often inconsistent in both direction and magnitude (; ). One reason is that bulk SOC integrates multiple organic matter fractions with contrasting turnover rates, accessibility, and persistence mechanisms. Therefore, bulk SOC alone is often insufficient for diagnosing short-term carbon dynamics or assessing management regulation (; ; Naipal et al., 2018).
Among soil organic matter fractions, dissolved organic matter (DOM) is one of the most mobile and biogeochemically reactive pools, and plays an important role in carbon cycling (; Ren et al., 2024). It originates mainly from litter leaching, root exudation, and microbial metabolites, and comprises a chemically diverse mixture of compounds, including humic-like and protein-like substances rich in reactive functional groups (e.g., carboxyl and phenolic hydroxyl moieties), which confer high reactivity during sorption, complexation, and biological transformation (; Speetjens et al., 2022). Owing to its mobility and compositional heterogeneity, DOM is closely involved in microbial metabolism, as well as transport and complexation of nutrients and contaminants. DOM also responds rapidly to shifts in vegetation input, soil properties, and management, making it particularly informative for tracing active carbon dynamics and profile-scale biogeochemical processes (; Schimel, 2010). Accordingly, DOM quantity and quality have been increasingly regarded not only as sensitive indicators of soil functional status and biogeochemical responses to environmental variation, but also as promising proxies for ecosystem functioning and, in some cases, persistence-related behavior of soil organic matter itself (; Tian et al., 2025).
Methodologically, DOM is widely characterized using ultraviolet–visible (UV–Vis) absorbance and fluorescence spectroscopy. Optical indices such as specific UV absorbance at 254 nm (SUVA254) and the absorbance slope ratio (SR) are commonly used to infer aromaticity and molecular-weight–related properties (Ohno, 2002; Weishaar et al., 2003). Fluorescence excitation–emission matrices (EEMs) combined with parallel factor analysis (PARAFAC) can identify humic-like and protein-like fluorescent components within DOM. Complementary indices (i.e., the fluorescence index (FI), humification index (HIX), and biological index (BIX)) further support interpretations of DOM sources and humification status (; Rinot et al., 2021). Therefore, these optical metrics enable integrated assessment of DOM composition, quality, and their underlying controls, thereby complementing bulk SOC measurements (; Teixeira et al., 2024).
Land use change can reconfigure DOM dynamics by reshaping vegetation composition, litter inputs, and the soil microenvironment. These effects may vary among restoration pathways and land-use transitions and may persist across the soil profile (; Weishaar et al., 2003). In dryland regions of the Loess Plateau, long-term cultivation has been widely replaced by or converted into grassland restoration, shrub establishment, forest development, and orchard planting (; Xu et al., 2018). These pathways may differ not only in the amount of organic carbon returned to soil, but also in the chemical nature and vertical distribution of organic inputs and their subsequent biological transformation into DOM (Liu et al., 2015). Nevertheless, most existing studies have emphasized broad differences among land use types, whereas the additional effects of vegetation characteristics and management practices within specific land use systems remain comparatively underexplored (; Zethof et al., 2019). For example, high-input versus low-input orchard management can alter litter return, soil disturbance, nutrient inputs, and surface crust development, thereby jointly regulating DOM quantity and quality (; Peng et al., 2023). In grassland systems, contrasting root architectures, such as taproot and fibrous-root systems, may influence root-derived carbon inputs and rhizosphere processing, leading to divergence in DOM composition and retention-related behavior (Peixoto et al., 2020; Peng et al., 2023; Wang et al., 2024). In shrublands, natural succession versus artificial planting can modify understory structure, resource competition, and shrub adaptation, with cascading consequences for organic inputs, rhizosphere processes, and soil properties, thereby altering DOM characteristics (Zhang et al., 2025). Therefore, beyond comparisons among land use types, incorporating variables within land use types, such as orchard management intensity, grassland root type, and shrubland restoration mode, can provide a more mechanistic understanding of DOM divergence during dryland restoration and land-use transitions (; Weishaar et al., 2003; Wilson and Xenopoulos, 2009).
Depth-dependent behavior is a defining feature of soil DOM. As DOM moves through the soil profile, its properties often change markedly with depth, and contrasts among soils may become more pronounced in deeper horizons. This means that sampling restricted to topsoil may underestimate subsoil physicochemical controls on DOM dynamics and function (; Tian et al., 2025). Biological soil crusts (BSCs), plant roots and litter influence DOM inputs and export by regulating aggregate stability, nutrient cycling, and near-surface hydrology, which in turn shape the redistribution of DOM from the soil surface into deeper layers of the soil profile (Riveras-Muñoz et al., 2022; Szyja et al., 2023). From a carbon-budget perspective, DOC leaching represents a recognized but often underestimated terrestrial carbon flux, whose magnitude and vertical expression are jointly regulated by hydrological transport and external inputs, resulting in divergent DOM responses between topsoil and subsoil and underscoring soil depth as a critical dimension for interpreting DOM change (; Nakhavali et al., 2020). Moreover, mineral-surface interactions and microbial transformation regulate DOM retention, release, and processing across mineral-associated and aggregate-protected domains, thereby shaping DOM quantity and quality along the profile (). However, most studies have focused either on topsoil or on vertical patterns within a single land use type, whereas integrated assessments that simultaneously consider land use type, restoration pathway, vegetation traits, soil depth, and management intensity within a unified field framework remain limited (; Li et al., 2019; Ren and Cai, 2025). Few field studies have combined a profile-scale comparison among land use types with embedded within-land-use contrasts under a common climatic and soil-forming setting. This limitation is particularly evident in drylands, where strong moisture limitation, deep rooting, and the juxtaposition of anthropogenic management and natural restoration can amplify profile-scale effects (; Young et al., 2021).
To address these gaps, we conducted a field comparison in a typical dryland watershed using 16 representative sites that captured broad land use settings together with embedded contrasts in vegetation and management. These contrasts included (1) shrubland under natural restoration (NR) versus artificially planted sites (AP); (2) grasslands dominated by taproot (TR) versus fibrous-root (FR) vegetation; and (3) orchards under high-input management (HM) versus low-input management (LM). Soils were sampled at three depth intervals (0–20, 30–50, and 60–80 cm), and optical techniques were used to characterize soil DOM across soil depth, land use, vegetation, and management, and to explore the underlying mechanisms. We hypothesized that: (i) Compared with cropland, restored woody and grassland systems would have higher DOC content and stronger humic-like and aromatic DOM signatures, reflecting greater organic matter inputs and reduced disturbance; (ii) DOM quantity and composition would show clear vertical differentiation along the soil profile, with declining DOC content and shifts in fluorescence composition with increasing depth; and (iii) vegetation traits, restoration mode, and management intensity would further reshape DOM characteristics within land use types, with distinct manifestations between topsoil and subsoil layers. Collectively, our findings would improve understanding of how dryland restoration pathway, soil depth, and within-system vegetation and management contrasts shape soil DOM dynamics, with implications for evaluating changes in active soil carbon pools under land-use transition.
2 Materials and methods
2.1 Study area overview and site selection
The study was conducted in the Zhifanggou watershed (8.27 km²) located in the typical hilly–gully region of the Chinese Loess Plateau. The watershed is highly dissected, with a gully density of 8.1 km km-² (; Zhao et al., 2016). The climate is typically dryland, with a mean annual temperature of 8.8 °C and mean annual precipitation of 505 mm. Precipitation is strongly seasonal and approximately 70% concentrated in June to September (Xu et al., 2018). Soils are dominated by young loess-derived soils developed on aeolian loess, typically silt loam in texture and classified as Typic Ustorthents according to USDA Soil Taxonomy (Messing et al., 2003). Organic matter in the cultivated horizon is generally low (0.53–0.77%) (Xu et al., 2018). Historically, inappropriate land use caused vegetation degradation, whereas soil and water conservation measures since the late 20th century have reshaped land use patterns. Currently, forestland, shrubland, and grassland dominate the watershed and together account for 83% of the area (). The main vegetation includes Robinia pseudoacacia, Populus simonii, Hippophae rhamnoides, Caragana korshinskii, and Artemisia sacrorum (; Zhang et al., 2026) (Figure 1).
Figure 1
Following a watershed-wide reconnaissance of topography and vegetation recovery, five land use types with comparable site conditions were selected from a pool of candidate sites. Each land use type comprised 2–4 vegetation sites differing in root architecture, management intensity, restoration pathway, and functional type, resulting in a total of 16 sites (Table 1). Orchard, grassland, shrubland, and forestland sites have experienced approximately 20 years of natural or assisted restoration, whereas sloping cropland remains under conventional local cultivation. Orchard management intensity was classified as a relative field category rather than as a strictly quantified input level. OMP site represented high-input management, characterized by more frequent field intervention and lower surface cover, whereas the OJR represented low-input management, with weaker visible disturbance and greater surface cover. Grassland sites comprised GBI and GSB as representative fibrous-root species, as well as GAS and GIS as representative taproot species. Shrubland sites included SCK and SHR representing long-term managed stands, together with SPS and SSM as typical naturally restored stands. Forestland sites were represented by FRP, FSA, and FPT. Cropland sites included CZM, CSI, and CST as representative crops (Table 1).
Table 1
| Land use type | Site code | Dominant species | Within-land-use category |
|---|---|---|---|
| Cropland | CZM | Zea mays L. | Crop species |
| CSI | Setaria italica (L.) Beauv. | Crop species | |
| CST | Solanum tuberosum L. | Crop species | |
| Orchard | OMP | Malus pumila Mill. | High-input management |
| OJR | Juglans regia L. | Low-input management | |
| Grassland | GBI | Bothriochloa ischaemum (L.) Keng | Fibrous-root |
| GSB | Stipa bungeana Trin. | Fibrous-root | |
| GAS | Artemisia sacrorum Ledeb. | Taproot | |
| GIS | Incarvillea sinensis Lam. | Taproot | |
| Shrubland | SCK | Caragana korshinskii Kom. | Artificially planted |
| SHR | Hippophae rhamnoides L. | Artificially planted | |
| SPS | Periploca sepium Bunge | Naturally restored | |
| SSM | Sophora moorcroftiana (Benth.) Baker | Naturally restored | |
| Forestland | FRP | Robinia pseudoacacia L. | Tree species |
| FSA | Salix babylonica L. | Tree species | |
| FPT | Pinus tabulaeformis Carr. | Tree species |
Land use type, site code, dominant species, and within-land-use comparison category for the 16 sampling sites.
2.2 Soil sampling and near-surface measurements
Across the five land use types, a total of sixteen vegetation sites were investigated. At each site, three 50 cm × 50 cm quadrats were established. At each quadrat, a soil pit was excavated, and approximately 0.5 kg of soil was collected from three depth intervals. The topsoil layer (0–20 cm) was strongly influenced by BSCs, litter inputs, and the main distribution of herbaceous roots. The middle soil layer (30–50 cm) was characterized by the presence of shrub roots, with residual effects of litter incorporation. The subsoil layer (60–80 cm) exhibited minimal litter influence and was penetrated only by a limited number of deep woody roots (; Wu et al., 2025). Samples from the three quadrats within each site were composited by depth, sealed in plastic bags, transported to the field station, and air-dried for subsequent analyses.
Near-surface attributes included biological soil crusts thickness (TBSC) and coverage (CBSC), plant root density (RD), litter density (LD), soil bulk density (BD), soil texture, and vegetation (including understory) cover. TBSC was defined as the thickness of the crust together with its adherent soil layer (Wang et al., 2017a, b; Yair et al., 2011) and was measured ten times using a caliper with 0.03-mm resolution following Wang et al. (2017a). CBSC was first estimated using the quadrat method (Liu et al., 2015) and subsequently refined by digital image analysis of color photographs, during which crust types were recorded (; Liu et al., 2015). Soil cores for RD and LD were collected using stainless-steel cylinders (10 cm diameter, 5 cm height). Roots and litter were separated by washing through a 1-mm sieve, oven-dried at 65 °C for 12 h, and averaged across sampling points (Wang et al., 2021). Vegetation cover was estimated by three observers in 1 m × 1 m quadrats, and the mean of their estimates was used (Wang et al., 2023a).
2.3 DOM extraction and DOC determination
Air-dried soils were gently crushed after removing visible plant residues and gravel, and then passed through a 1-mm sieve to homogenize the material. A total of 48 soil samples were obtained (16 sites × 3 depths), and each sample was processed in quadruplicate. Water-extractable DOM was prepared using ultrapure water. Briefly, 10 g of sieved soil was combined with 100 mL ultrapure water (10:1, v/w) in glass Erlenmeyer flasks, sealed with Parafilm, and shaken at 180 rpm for 60 min at 40 °C. The resulting slurry was centrifuged at 6000 rpm for 6 min, after which the supernatant was filtered through a 0.45-µm membrane to obtain the DOM extract (; Li et al., 2019).
To limit photodegradation, the 192 extracts were split into two aliquots and stored in amber vials at 4 °C in the dark. DOC and optical analyses (UV–Vis absorbance and EEM fluorescence) were completed as soon as possible. DOC was measured using a TOC analyzer (TOC-L; Shimadzu, Japan) (Li et al., 2019). Because carbon typically accounts for approximately 67% of organic matter on an elemental basis, DOC was used as a quantitative proxy for DOM concentration in this study (; Li et al., 2019).
2.4 Spectroscopic analyses
2.4.1 Fluorescence EEMs and PARAFAC modeling
EEMs were acquired on a fluorescence spectrophotometer (RF-6000; Shimadzu, Japan). Excitation wavelengths were scanned from 200 to 500 nm and emission was recorded from 250 to 550 nm, with step sizes of 5 nm (Ex) and 2 nm (Em); the scan speed was 2000 nm min-¹. Prior to measurement, extracts were diluted with deionized water to bring DOC below 8 mg L-¹, thereby reducing inner-filter artifacts (; ). Raw EEMs were processed by subtracting an ultrapure-water blank to correct Raman scatter, masking Rayleigh scattering regions (set to zero), and applying inner-filter correction when necessary, using the corresponding absorbance spectra (; Ohno and Bro, 2006).
PARAFAC modeling was performed in MATLAB using the DOMFluor toolbox (Stedmon and Bro, 2008). Candidate models containing two to seven components were evaluated using loading and residual diagnostics. Split-half analysis was performed with two independent split combinations, and random-initialization analysis used 10 starts. Component congruence between the full-data and split-half solutions was examined using Tucker congruence coefficients. The three-component solution was retained because it was validated in both split combinations and produced stable, interpretable excitation and emission loadings.
Three fluorescence indices were calculated from the corrected EEMs. FI was defined as the emission intensity ratio at 470 nm to 520 nm at Ex = 370 nm, reflecting the relative contributions of autochthonous and allochthonous sources of DOM (). HIX was computed as the ratio of the integrated emission area over 435–480 nm to that over 300–345 nm at Ex = 254 nm and served as an indicator of humification (Ohno, 2002; Wilson and Xenopoulos, 2009). BIX was calculated as the emission ratio at 380 nm to 430 nm at Ex = 310 nm, indicating autochthonous and freshly produced DOM signatures ().
2.4.2 UV–visible absorbance spectroscopy
UV–Vis absorbance spectra were recorded with a UV-Vis spectrometer (UV-1780; Shimadzu, Japan) using a 10 mm quartz cuvette and ultrapure water as the blank. Absorbance was scanned from 200 to 400 nm at 1 nm intervals (Shafiquzzaman et al., 2014). Specific UV absorbance at 254 nm (SUVA254; L mg-¹ m-¹) was calculated by normalizing absorbance at 254 nm to DOC to indicate aromaticity per unit carbon (Lv et al., 2023; Weishaar et al., 2003). The spectral slope ratio (SR), derived from slopes over 275–295 nm and 350–400 nm, was used as a proxy for relative molecular weight (; ).
2.5 Statistical analyses
All variables were tested for normality and homogeneity of variances prior to inferential analyses. When assumptions were met, one-way ANOVA was used to assess differences in DOC content and optical metrics among sites or among depth intervals. Significant effects (P < 0.05) were followed by post hoc multiple comparisons using Duncan tests. For paired contrasts among vegetation–management factors within the same depth, paired-sample t-tests were conducted. Unless otherwise stated, numerical values are reported as mean ± standard deviation (SD), and all figure error bars denote SD.
Pearson correlation analysis and principal component analysis (PCA) were used to summarize co-variation among DOC content, UV–Vis metrics, fluorescence indices, and PARAFAC components. These analyses were used to explore associations between DOM properties and soil physicochemical or near-surface attributes (e.g., texture, BD, RD, LD, vegetation cover, and BSC traits), rather than to infer causal drivers or independent explanatory strength. Before PCA, all included variables were centered and scaled to unit variance. PCA used site-depth means for all 16 sites at the three depths (n = 48), whereas Pearson correlations with surface attributes were restricted to the 16 topsoil observations. All statistical analyses were performed in SPSS 26.0 (SPSS Inc., Chicago, IL, USA), and figures were prepared using Origin 2025b (OriginLab, Massachusetts, USA).
3 Results
3.1 Variation in soil DOC content
Forestland generally exhibited the highest DOC content, followed by shrubland and grassland, whereas orchard and cropland consistently showed the lowest values. Along the soil profile, DOC content generally declined with depth across all land use types, with significantly higher values in the topsoil than in the middle and subsoil layers. Additional variation was observed among vegetation and management contrasts within several land use types (Figure 2).
Figure 2
This overall land-use pattern persisted across the soil profile. In forestland, DOC content at each depth, declining from 235.22 ± 87.05 mg kg−1 in the topsoil to 114.47 ± 44.80 and 62.32 ± 31.36 mg kg−1 in the middle and subsoil layers, respectively. Orchard and cropland remained lowest (10.16–89.15 mg kg−1 in topsoil and generally <30 mg kg−1 at depth). Forestland exceeded the other land use types at all depths (P < 0.05). Shrubland, grassland, and orchard did not differ in topsoil, but shrubland exceeded grassland and orchard in the deeper layers (Figure 2b).
Further differences emerged within individual land use types. The walnut/low-input orchard site had consistently higher DOC content than the apple/high-input orchard site across all depths (P < 0.05). In cropland, maize had lower DOC content than millet and potato, including a topsoil value of 10.16 ± 4.34 mg kg−1. Differences were smaller within shrubland and grassland (Figure 2a).
3.2 Fluorescence components of soil DOM
EEM–PARAFAC modeling resolved three fluorescence components across land use types and soil depths. Component C1 was assigned to a UVA humic-like fluorophore, typically associated with higher-molecular-weight and hydrophobic-acid–like fractions (Wang et al., 2023b). Component C2 was commonly regarded as a terrestrial humic-like component (Meilleur et al., 2023). Component C3 represented protein-like material, characterized by tyrosine- and tryptophan-like optical signatures (Williams et al., 2013) (Figure 3).
Figure 3
In the topsoil, humic-like components (C1 + C2) accounted for more than 70% of total fluorescence across all sites, whereas the protein-like component (C3) contributed a smaller proportion. Cropland and grassland generally showed higher C3 proportions than shrubland, forestland, and orchards, with orchards exhibiting the lowest C3 contribution. In addition, high-input orchards showed a higher humic-like proportion than low-input orchards, whereas root type and restoration mode had little effect on component proportions (Figures 4a, b).
Figure 4
In deeper layers, humic-like components still accounted for more than 50% of total fluorescence, but the protein-like fraction increased, especially in shrubland and cropland. Shrubland showed the highest C3 proportion, whereas orchards remained lowest; within cropland, potato sites had an approximately twofold higher C3 proportion than millet and a 3.5-fold higher proportion than maize. In the subsoil, the shrubland C3 proportion further increased to 49%, whereas the effect of orchard management became negligible (Figures 4c-f).
Vertical differences in component proportions were weak in cropland, where only C2 was higher in the topsoil than in the middle layer (P < 0.05). Orchards showed higher topsoil C2 but higher subsoil C3, whereas shrubland and forestland exhibited a clear decrease in C1 and C2 and a corresponding increase in C3 with depth. Overall, C1 and C2 were significantly enriched in topsoil, whereas C3 increased toward deeper layers (P < 0.001; Figures 5a-c).
Figure 5
3.3 Excitation emission matrix fluorescence indices of soil DOM
3.3.1 Variation in FI and BIX of soil DOM
FI and BIX showed similar land-use and depth patterns, although FI differences were less pronounced. Cropland had the highest values, followed by grassland and orchard, whereas shrubland and forestland had the lowest. Both indices generally increased with depth, with topsoil values lower than those in the middle and subsoil layers (P < 0.05; Figures 6a, 7a).
Figure 6
Figure 7
Depth-specific contrasts among land use types were evident for FI and BIX values, particularly for BIX values. In cropland, subsoil BIX values reached 1.85 ± 0.13, decreasing to 1.80 ± 0.13 and 1.65 ± 0.12 in the middle and topsoil layers, respectively. Grassland and orchard exhibited the next highest BIX values, reaching 1.59–1.63 in the subsoil, whereas BIX values in the topsoil and middle soil layer generally did not exceed 1.45. By contrast, shrubland and forestland were characterized by low BIX values, with subsoil BIX values of only 1.19–1.35 and topsoil typically below 0.8. Across all land use types, topsoil BIX values were significantly lower than those in the middle and subsoil layers (P < 0.05) (Figure 7b). Moreover, across soil layers, FI and BIX values generally followed the order of cropland > grassland > orchard > shrubland > forestland (Figures 6b, 7b).
Within a given land use type, FI and BIX values also varied among vegetation types. In orchards, BIX values under high-input management were significantly higher than under low-input management (P < 0.05). In grassland, FI values were significantly higher in taproot sites than in fibrous-root sites (P < 0.05). In shrubland, FI values were significantly higher in artificially restored sites than in naturally successional sites (Figures 6a, 7a).
3.3.2 Variation in HIX of soil DOM
HIX values differed markedly among land use types and soil depths. Overall, forestland exhibited the highest HIX values, followed by shrubland and orchard, whereas grassland and cropland had the lowest HIX values. Along the soil profile, HIX values declined with depth across all land use types, with significantly higher HIX values in the topsoil than in the middle and subsoil layers (P < 0.05; Figure 8a). Moreover, within each land use type, HIX varied among vegetation types with contrasting traits and management regimes (Figure 8a).
Figure 8
The depth-specific differences among land use types were pronounced. In forestland, HIX values in the topsoil reached 11.14 ± 3.93 and decreased to 5.77 ± 0.36 and 5.61 ± 0.44 in the middle and subsoil layers, respectively. In shrubland, HIX values decreased from 10.01 ± 4.41 in the topsoil to 4.59 ± 0.53 and 3.87 ± 0.45 in the middle and subsoil layers. In orchards, topsoil HIX values reached 7.80 ± 0.77, whereas middle- and subsoil remained below 3.5. Grassland and cropland exhibited consistently low HIX values, with topsoil values ranging from 3.10 to 6.53 and subsoil values generally below 2.1. Across land use types, topsoil HIX values were significantly higher than those in the middle and subsoil layers (P < 0.05), whereas differences between the middle and subsoil layers were significant only at some sites (Figure 8b). Furthermore, across soil layers, HIX generally followed the order of forestland > shrubland > orchard > grassland > cropland. However, differences between forestland and shrubland, and between orchard and grassland, were not significant in the topsoil but became significant in the middle and subsoil layers (P < 0.05; Figure 8b).
Within land use types, vegetation-related contrasts in HIX values were most evident in the topsoil. In forestland, topsoil HIX values ranged from 8.39 to 15.56 across tree species, while the ranges were much narrower in the middle (5.54–5.88) and subsoil layers (5.27–5.77). Willow exhibited a significantly higher topsoil HIX than other tree species (P < 0.05), but this difference diminished with increasing depth. In orchards, topsoil HIX values differed significantly between management regimes, with higher values under low-input management than under high-input management (P < 0.05). Notably, shrubland and grassland also showed significant vegetation-type differences in topsoil HIX values, but these differences were not associated with root type or restoration mode. In shrubland, SCK and SPS exhibited significantly higher HIX values than the other shrub species in the topsoil (P < 0.05), while in grassland, GSB and GIS had higher HIX values than the other grassland types in the topsoil (P < 0.05). By contrast, cropland showed no clear differences among crops across depths. Overall, vegetation-type effects within a land use category were strongest in the topsoil and weakened with depth (Figure 8a).
3.4 UV–visible spectral indices of soil DOM
The SR and SUVA254 values differed significantly among land use types and across soil depths, and the two indices showed opposite trends. Overall, forestland exhibited the highest SUVA254 values and the lowest SR values, followed by shrubland and grassland, whereas orchard and cropland showed the lowest SUVA254 values and the highest SR values. Along the soil profile, SR generally increased with depth whereas SUVA254 decreased with depth across land use types (Figures 9a, 10a). Meanwhile, within each land use type, SR and SUVA254 values varied among vegetation types with contrasting traits and management regimes (Figures 9a, 10a).
Figure 9
Figure 10
Depth-specific contrasts among land use types were clear. In cropland, topsoil SR values reached 2.27 ± 0.88, decreasing to 1.88 ± 0.36 in the middle soil layer and increasing again to 2.23 ± 0.20 in the subsoil; meanwhile, SUVA254 values remained below 1.1 across all depths. In orchards and grassland, subsoil SR values ranged from 1.95 to 2.1, whereas SR in the topsoil and middle soil layer declined to 1.3–1.6; SUVA254 values reached 2.5–3.4 in the topsoil and remained below 2 in the middle and subsoil layers. In shrubland and forestland, SR values were consistently <1.5 across depths, whereas SUVA254 reached 3.4–4.3 in the topsoil and remained ≥2 in deeper layers. Except for cropland, subsoil SR values were significantly higher than those in the topsoil and middle soil layers (P < 0.05), while SUVA254 values were significantly higher in the topsoil than in the middle and subsoil layers (Figures 9b, 10b). Furthermore, across soil layers, SR values generally followed the order of cropland > orchard and grassland > shrubland > forestland. In the middle and subsoil layers, SUVA254 values generally followed forestland and shrubland > grassland > orchard and cropland. In addition, differences among land use types became less pronounced with increasing depth (Figures 9b, 10b).
Within a given land use type, SR and SUVA254 also varied among vegetation types and management or restoration regimes. In forestland, willow exhibited significantly higher SUVA254 values (5.44) than the other two tree types (P < 0.05). In orchards, low-input management was associated with lower SR in the topsoil but higher SUVA254 in both the topsoil and subsoil compared with high-input management (P < 0.05). In grassland, subsoil SR was significantly higher in taproot sites than in fibrous-root sites (P < 0.05). In shrubland, SUVA254 in the middle and subsoil layers was significantly higher under artificially restored sites than in naturally successional sites (P < 0.05). In cropland, maize and the other two crops differed in their vertical patterns. Maize exhibited the lowest SR values and the highest SUVA254 values in the topsoil, whereas this feature occurred in the middle soil layer for the other crops (Figures 9a, 10a).
3.5 Principal component analysis and correlation analysis of soil DOM
PCA was used to visualize multivariate variation among the 48 site-depth observations in relation to land use type and soil depth. The first two principal components (PC1 and PC2) explained 34.3% and 24.6% of the total variance, respectively. The three depth groups overlapped but showed a gradient along PC1, with topsoil observations tending toward positive scores and middle- and subsoil observations shifting toward negative scores. PC1 was positively correlated with DOC, SUVA254, HIX, components C1 and C2, LD, and the fine-texture fractions (clay and silt), but negatively correlated with FI, BIX, SR, BD, RD, component C3, and sand content. Vector geometry further indicated small angles between LD, SUVA254, DOC, and clay, as well as between BD, SR, FI, and sand. In addition, C1 and C2 were positioned close to HIX, whereas C3 was positioned close to BIX (Figure 11).
Figure 11
Pearson correlation analysis (Figure 12) further supported these patterns for the topsoil. SR was strongly positively correlated with BIX and BD (P < 0.01), but negatively correlated with SUVA254, RD, and LD (P < 0.01). SUVA254 was strongly positively correlated with DOC, C2, RD, and LD (P < 0.01), and negatively correlated with SR, FI, and BD (P < 0.01); FI showed correlation patterns similar to SR. HIX was strongly positively correlated with C1 and negatively correlated with C3 (P < 0.01). BD was strongly negatively correlated with RD, LD, TBSC, and clay content (P < 0.01). For surface properties, TBSC was strongly positively correlated with silt content and CBSC, but negatively correlated with BD and sand content (P < 0.01, Figure 12).
Figure 12
4 Discussion
4.1 Spatial variation in soil DOC content
In this study, differences in DOC content among land use types were most evident in the topsoil (Figure 2), with higher values in forestland and shrubland, followed by grassland, and lower values in orchards and cropland. This pattern was associated mainly with differences in surface organic inputs and near-surface processes (; Sun et al., 2026).
At the land use scale, differences in DOC content can be explained by three aspects, which include organic inputs, soil microenvironment, and physical retention (; Zhang et al., 2025). First, variation in organic inputs was strongly associated with topsoil DOC. Forestland had the highest DOC content, consistent with its higher litter density (LD), root density (RD), and biological soil crust thickness (TBSC) (Supplementary Figure 1). In contrast, cropland had relatively low LD and RD across depths (Supplementary Figures 1a-c). PCA showed a positive association between DOC and litter density, consistent with an important role of surface organic inputs in DOC variability (Figure 11). Second, land use influenced DOC retention by altering soil texture and structure. Cropland had a lower clay content but a higher sand content (Supplementary Figures 1a-c), which may promote DOC loss via infiltration and leaching and reduced retention, consistent with the mechanism proposed by . Third, land use effects on soil temperature and moisture were not measured here, but published studies indicate that these conditions can influence DOC production and transport (; Young et al., 2021). For example, higher water-holding capacity in grassland and deeper soil drying in forestland in arid and semiarid regions may alter DOC transport along the profile, although this effect mainly modified vertical patterns rather than driving overall land use differences (; van Dijk and Keenan, 2007).
DOC content decreased significantly with increasing soil depth across all land use types (P < 0.05), although the magnitude of decline varied. This depth-related decline is consistent with the surface enrichment of LD and the progressive weakening of fresh organic inputs with depth (Supplementary Figure 1). In deeper layers, DOM is sustained mainly by limited deep-root inputs or the vertical transport of DOM from the surface, while mineral sorption and continuous microbial consumption further reduce the DOC pool (; Teixeira et al., 2024). Importantly, the decline rate varied among land use types. In the topsoil, orchards, grassland, and shrubland had relatively high surface inputs, which partly explains why their DOC content could overlap in the topsoil. In the middle and subsoil layers, however, shrubland maintained higher DOC content than grassland and orchards (Figure 2b), likely due to deeper carbon inputs from woody roots. Forestland showed the steepest decline with depth, consistent with strong surface litter and root inputs but a rapid reduction in carbon supply and microbial activity with depth (Wang et al., 2024). Grassland exhibited a more gradual decrease, consistent with relatively high root density and well-developed biological soil crusts that support DOC turnover and retention (Supplementary Figure 1) (Riveras-Muñoz et al., 2022; Szyja et al., 2023).
Within land use types, management and vegetation types further modified DOC content. In orchards, the combined species-management contrast was associated with DOC content, with consistently lower values at the apple/high-input site than at the walnut/low-input site across depths (P < 0.05; Figure 2b). The observed difference may reflect greater surface cover and carbon retention at the walnut/low-input site (Wang et al., 2024) and stronger disturbance at the apple/high-input site, but species and management effects could not be separated in this field comparison. In grassland, contrasting root types can alter root distribution and rhizosphere inputs (Peixoto et al., 2020; Teixeira et al., 2024), yet DOC profiles did not differ clearly between root systems. This suggests that root-derived inputs may be reshaped by subsequent leaching, microbial processing, and mineral sorption (; Nakhavali et al., 2020). In shrubland, artificially planted sites (AP) showed higher RD and higher understory vegetation cover than naturally restored sites (NR) (P < 0.05; Supplementary Figure 1), but DOC did not differ significantly between AP and NR. This suggests that greater organic inputs in AP may have been partly offset by faster microbial processing and DOC consumption (Wang et al., 2023a).
4.2 Spatial variation in DOM fluorescence components
At the land use scale, fluorescence component patterns reflected how vegetation type and disturbance intensity reshape DOM sources. Across all land use types, humic-like components dominated the total fluorescence signal (Figure 4), which was consistent with a substantial contribution from plant-derived material. This agrees with observations from dryland sites in Israel, where DOM was largely controlled by plant residues and vegetation-derived organic matter, with optical indicators pointing to strong plant-derived and humic-like substances (Nasonova et al., 2022; Rinot et al., 2021). Although humic-like components were dominant everywhere, the protein-like fraction tended to be relatively higher in cropland and grassland and to increase strongly in shrubland with depth, while orchards generally exhibited the lowest protein-like contribution (Figure 4). Such differences likely reflect contrasts in organic input quality and microbial processing. Forestland and shrubland typically provide lignin- and polyphenol-rich litter and more stable organic inputs, which can be transformed by microbes into larger and more aromatic humic-like DOM, leading to higher humic-like signatures (; ; ). Supporting this, humic-like components were positively related to RD, and land use types with higher RD tended to show stronger humic-like substances (Supplementary Figure 1). This highlights root systems as an important link between land use and DOM composition, because higher RD can enhance rhizosphere carbon inputs and microbial transformation, promoting humic-like DOM formation. Similar patterns have been reported on the Loess Plateau, where forestland DOM showed higher aromaticity and hydrophobicity as well as stronger humic-like features than cropland (Liu et al., 2015).
Vertically, the topsoil was generally enriched in humic-like substances, while the relative contribution of protein-like components increased with depth, suggesting a shift toward more microbially processed or fresh characteristics in deeper DOM (Meng et al., 2024; Wang et al., 2024). This depth trend is consistent with Loess Plateau profile studies showing declining humic-like substances and increasing protein-like substances with profile (Li et al., 2019; Meilleur et al., 2023). Except for cropland, the protein-like component was relatively low in the topsoil, which may reflect long-term microbial assimilation of small protein-like molecules and their conversion into larger humic-like substances during sustained revegetation. The strong negative relationships between C3 and both C1 and C2 (P < 0.001) further indicate contrasting distribution patterns between protein-like and humic-like components, consistent with differential processing of these fluorophores along the soil profile (; Ren and Cai, 2025). In addition, stronger litter inputs in the topsoil (Supplementary Figure 1) promote humic-like formation and stabilization through mineral adsorption and complexation, resulting in significantly higher C1 and C2 contributions in the topsoil than in deeper layers (P < 0.05) (Meilleur et al., 2023; Williams et al., 2013). With increasing depth, decreasing C1 and C2 contributions suggest that deep DOM is not a simple extension of surface DOM into deeper layers. Instead, deep root activity and microbial metabolism under substrate limitation may generate more microbially derived soluble products, increasing the relative protein-like signal (; ).
Within land use types, management and restoration factors acted mainly as modifiers of component proportions. In orchards, high-input management at the apple site involved more frequent disturbance and reduced litter return, which may have contributed to lower protein-like fluorescence than at the walnut/low-input site; because tree species and management co-varied, their effects could not be separated. In shrubland, artificially planted sites often had higher understory coverage and higher LD, providing more terrestrial organic inputs and thus a higher humic-like component in the topsoil. With increasing depth, differences in surface inputs weakened, and the component proportions of AP and NR tended to converge (Figure 4). This pattern is consistent with the idea that deep DOM is mainly controlled by transport, mineral interactions, and limited inputs within the subsoil (; ).
4.3 Spatial variation in fluorescence spectral indices of soil DOM
Fluorescence indices provided optical evidence related to DOM processing and humification (; Ohno, 2002). FI varied little among land use types (Figure 6), whereas BIX and HIX showed clearer contrasts. Higher BIX in cropland and grassland was consistent with stronger microbially associated or freshly produced DOM signatures, whereas higher HIX and lower BIX in forestland and shrubland were consistent with more humified and plant-associated optical signatures (; ). FI, BIX, and fluorescence-component assignments are proxies influenced by source, processing history, and overlapping fluorophores and therefore do not uniquely identify DOM origin.
These patterns can be linked to disturbance intensity and substrate quality. In cropland, frequent disturbance was associated with higher bulk density (BD) and lower LD and RD (Supplementary Figure 1), favoring rapid microbial cycling of labile substrates. Under such conditions, microbial metabolites can contribute a larger fraction of DOM, which could contribute to higher FI and BIX and lower HIX. In forestland and shrubland, weaker disturbance and larger inputs of woody litter may promote repeated processing and accumulation of humic-like substances, resulting in higher HIX and lower FI and BIX (Zethof et al., 2019). Soil structure and near-surface features can further regulate these indices. In cropland, high BD and a constrained pore structure may promote rapid cycles of DOM production and consumption and thereby strengthen microbial signals. In contrast, in forestland and grassland topsoil, more stable soil structure and stronger BSCs development (Supplementary Figure 1) can enhance physical protection of DOM on mineral surfaces or within aggregates, allowing more extensive processing and humification (; Liu et al., 2015; Wilson and Xenopoulos, 2009).
A second noteworthy pattern was that significant vegetation-type differences in topsoil indices, especially HIX, appeared in grassland and shrubland; but these differences did not correspond clearly to root type or restoration mode. This suggests that site-to-site factors, such as microbial activity, litter chemistry, or microhabitat conditions, may outweigh the single-factor contrasts defined in our design. Higher microbial activity, for instance, could promote repeated processing and thus enhance humic substances, leading to higher HIX in certain vegetation types (Zethof et al., 2019).
Across all land use types, DOM generally shifted from a more humified and aromatic state in the topsoil to a deeper DOM pool with stronger microbially associated optical signatures and smaller apparent molecular size. This vertical differentiation does not mean that deep DOM is simpler, but rather it reflects stronger environmental constraints on DOM formation in deeper horizons. Compared with topsoil, the source structure of deep DOM likely depended more on vertical transport of soluble carbon and deep-root inputs, rather than on direct inputs of plant residues (; Szyja et al., 2023). The PCA also supports this interpretation (Figure 11). Mechanistically, topsoil receives concentrated litter and rhizosphere inputs and may experience more favorable moisture or temperature conditions for repeated decomposition and re-synthesis, promoting humic-like DOM and higher HIX. In deeper layers, limited carbon and energy supply, together with selective retention of large and aromatic DOM during transport, can increase the relative intensity of microbially associated fluorescence and raise FI and BIX while lowering HIX (; ).
Within land use types, species-management contrasts were associated with the balance between fresh and humified substances, but the effects were often weaker than the main land use contrast. In orchards, the apple/high-input and walnut/low-input sites differed in BIX and HIX. The apple/high-input site showed stronger fresh, microbially associated signatures, whereas the walnut/low-input site showed stronger humification (Figures 7 and 8). Intensive orchard management can reduce humic substances by lowering surface organic matter accumulation and increasing disturbance (; Liu et al., 2015; Messing et al., 2003). However, the observed differences should not be attributed to management alone because tree species, litter chemistry, and management-related changes in soil properties can jointly influence microbial processing and mineral stabilization. In grassland and shrubland, root type and restoration mode showed relatively weak effects on FI, BIX, and HIX. A likely reason is that newly produced DOM can be quickly consumed or stabilized by minerals and aggregates, which can reduce persistent differences in optical indices at the profile scale. Similarly, deeper root inputs may strengthen microbial contribution locally, but these effects may be weakened when DOM is integrated across soil horizons and transport pathways (Rinot et al., 2021; Witzgall et al., 2024).
4.4 Spatial variation in UV–Visible spectral indices of soil DOM
Across land use types, SR and SUVA254 exhibited opposite patterns, with higher SR but lower SUVA254 in cropland and grassland, and the reverse in shrubland and forestland (Figures 9b, 10b). Lower SR in woody systems was consistent with more humified DOM of larger apparent molecular size, whereas higher SR in cropland and grassland was consistent with smaller apparent molecular size and more microbially associated optical signatures. This interpretation is consistent with the component-based results, which show relatively higher protein-like contributions in cropland and grassland, whereas humic-like components dominate in woody systems (Figure 4). Woody litter inputs typically contain more lignin and polyphenols; after microbial processing, these inputs more readily form aromatic, humic-like DOM, leading to higher SUVA254 and lower SR. Conversely, crop residues and root inputs often contain more easily decomposed carbohydrates and lower-aromatic substrates, which may yield optical signatures of lower aromaticity and smaller apparent molecular size. Frequent disturbance may also weaken consistent depth contrasts in cropland (Weishaar et al., 2003; Zethof et al., 2019).
Along soil profiles, SR generally increased and SUVA254 generally decreased with depth across land use types. Together with the PCA and correlation results (Figures 11, 12), this pattern indicates that deep DOM differed strongly from topsoil DOM. Because LD was concentrated near the surface (Supplementary Figure 1), deep DOM likely depended less on subsoil litter inputs and more on vertical transport from the surface and limited deep-root inputs. During transport, DOM fractions with optical signatures of greater aromaticity and larger apparent molecular size may be preferentially retained through mineral adsorption, leaving deeper DOM with smaller apparent molecular size and weaker aromatic optical signatures (; ; Rinot et al., 2021). However, depth responses were not strictly linear across all land uses. Differences in soil structure stability and root input patterns may change the magnitude of these shifts (Sae-Tun et al., 2023; Witzgall et al., 2024). For example, in forestland and shrubland, deep roots and relatively stable soil structure may partly buffer the decline in aromaticity with depth, whereas in cropland and orchards, stronger disturbance and lower organic inputs may promote a shift toward optical signatures of smaller apparent molecular size and lower aromaticity in deeper layers (Nakhavali et al., 2020; Zethof et al., 2019).
A notable exception was observed in some cropland sites such as millet and potato, where topsoil SUVA254 was relatively low but SR was high, consistent with DOM enriched in low-aromatic fractions of smaller apparent molecular size. One likely explanation is that tillage and harvesting disturbances break aggregates and increase soil aeration, exposing previously protected organic matter to rapid decomposition (Weidhuner et al., 2021). At the same time, root exudates and microbial metabolites can accumulate near the surface and turn over quickly, and these products commonly exhibit low-aromatic optical signatures and small apparent molecular size. In contrast, more aromatic humic-like DOM may be selectively adsorbed and retained by clay and reactive minerals in the middle soil layer, causing a relative enrichment of aromatic structures and a local increase in SUVA254 at that depth (; ).
The orchard species-m anagement contrast was also associated with DOM optical structure. In orchards, the apple/high-input site generally showed weaker aromatic optical signatures than the walnut/low-input site (P < 0.05), together with a smaller apparent molecular size in the topsoil, although not always significant (Figures 9a; 10a). These differences may reflect stronger disturbance and lower surface cover at the apple/high-input site, together with species-specific litter chemistry (Supplementary Figure 1d), which may shorten DOM residence time in the topsoil and enhance vertical transport or rapid microbial consumption, leaving DOM with weaker aromatic signatures and smaller apparent molecular size. Walnut orchards, with less disturbance and more surface cover, may allow repeated microbial processing and stronger mineral interactions in the topsoil, resulting in stronger aromatic signatures and larger apparent molecular size (; ). In grassland, we only detected a clear root-type effect in the subsoil, where taproot dominated sites showed higher SR than fibrous-root sites (P < 0.05; Figure 9a), suggesting that deeper rooting may deliver more labile, smaller dissolved carbon compounds into deeper soil layers (Messing et al., 2003). In shrubland, artificially planted sites showed higher aromaticity than naturally restored sites in the middle and subsoil layers (P < 0.05; Figure 10a), while topsoil differences were small. This implies that artificial restoration may improve mid–deep soil conditions and promote microbial processing and stabilization of DOM, whereas in the topsoil, both restoration modes may converge due to generally high surface cover and BSCs development (; Riveras-Muñoz et al., 2022; Szyja et al., 2023).
Taken together, hypotheses (i) and (ii) were broadly supported: restored woody and grassland systems generally differed from cropland in DOC and optical composition, and DOM showed strong vertical differentiation. Hypothesis (iii) was partly supported. The orchard species-management contrast was clear for several indicators, whereas grassland root type and shrubland restoration mode produced weaker and indicator-dependent responses.
4.5 Study limitations and implications
This study establishes a field framework for interpreting soil DOM dynamics in drylands by integrating land use type, soil depth, and selected vegetation management contrasts. Unlike conventional land use comparisons, our design tracks DOM from topsoil to deeper layers while explicitly contrasting orchard management intensity, grassland root type, and shrubland restoration mode within the same sampling scheme. Using a consistent set of optical and fluorescence indicators and relating them to surface organic inputs and soil physical properties, we identify where land use effects are strongest along the profile and which factors most consistently explain DOM differences. These findings provide a practical basis for evaluating restoration effectiveness and optimizing dryland management by tracking changes in active carbon pools rather than relying only on bulk soil carbon.
Some limitations should be acknowledged. The study was conducted at representative sites with similar climate, topography, and soil-forming conditions, which helped reduce background variability but may limit extrapolation to other environmental settings. Because the sites represented existing field systems rather than a fully factorial experiment, species identity, vegetation traits, and management categories covaried; in particular, the apple/high-input and walnut/low-input orchards could not isolate management effects independently of tree species. These design features limit independent attribution of individual factors. In addition, interpretations were based mainly on optical and fluorescence indicators, which describe DOM composition but do not directly resolve the microbial and mineral processes controlling DOM transformation and stabilization. Moreover, DOM in dryland soils responds strongly to rainfall pulses, whereas the present sampling represents conditions within a limited time period and does not capture short-term hydrological effects.
Future work should extend observations across broader environmental gradients and integrate measurements of microbial activity, mineral composition, and soil structural properties. Seasonal sampling and event-based monitoring would further clarify DOM responses to rainfall variability and strengthen process-based interpretation. Such efforts will improve the use of DOM indicators in evaluating ecological restoration and land management in dryland systems.
5 Conclusion
Using DOC measurements, UV–Vis spectroscopy, EEM–PARAFAC, and multivariate analysis, this study examined profile-scale DOM across five land-use types and within-type contrasts in a typical dryland watershed of the hilly–gully Loess Plateau. Land use and soil depth structured DOM most strongly: forestland and shrubland had higher DOC contents and stronger humic-like and aromatic optical signatures, whereas deeper soils showed lower DOC contents and optical signatures consistent with greater microbial contributions. The coupled orchard tree species–management contrast was evident, whereas the effects of grassland root type and shrubland restoration mode were weaker and indicator-dependent.
These findings indicate that revegetation and subsequent management are associated with shifts in the quantity, composition, and vertical distribution of an active and mobile soil carbon pool. Maintaining surface cover and organic inputs while limiting unnecessary disturbance may help sustain DOM and humic-like optical characteristics during ecological restoration on the Loess Plateau, although this implication remains context-dependent. The conclusions most directly represent the investigated watershed and may inform comparable dryland settings. Future seasonal and event-based observations, together with microbial, mineralogical, and hydrological measurements, are needed to test their wider transferability.
Statements
Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://github.com/wangyicong031225-Lando/Research-data.git.
Author contributions
Y-CW: Data curation, Formal Analysis, Investigation, Visualization, Writing – original draft. Q-WZ: Conceptualization, Resources, Supervision, Writing – review & editing, Methodology, Validation. N-YL: Data curation, Visualization, Writing – original draft. Y-HZ: Investigation, Writing – original draft, Visualization. GC: Data curation, Investigation, Writing – original draft. D-XW: Data curation, Writing – original draft, Formal Analysis. G-XG: Writing – original draft, Data curation, Visualization. QZ: Investigation, Writing – original draft, Formal Analysis. F-BZ: Investigation, Writing – review & editing. JW: Resources, Writing – review & editing, Methodology. ML: Resources, Writing – review & editing, Methodology, Software, Validation. HW: Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing.
Funding
The author(s) declared that financial support was received for this work and/or its publication. This work was funded by the National Natural Science Foundation of China (42477362).
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2026.1934903/full#supplementary-material
References
1
AikenG. R.Hsu-KimH.RyanJ. N. (2011). Influence of dissolved organic matter on the environmental fate of metals, nanoparticles, and colloids. Environ. Sci. Technol.45, 3196–3201. doi: 10.1021/es103992s
2
BeillouinD.CorbeelsM.DemenoisJ.BerreD.BoyerA.FallotA.et al. (2023). A global meta-analysis of soil organic carbon in the Anthropocene. Nat. Commun.14, 3700. doi: 10.1038/s41467-023-39338-z
3
BolanN. S.AdrianoD. C.KunhikrishnanA.JamesT.McDowellR.SenesiN. (2011). “ Chapter one - Dissolved organic matter: Biogeochemistry, dynamics, and environmental significance in soils,” in Advances in Agronomy. Ed. SparksD. L. (San Diego: Academic Press), 1–75. doi: 10.1016/B978-0-12-385531-2.00001-3
4
ChamizoS.Rodríguez-CaballeroE.RománJ. R.CantónY. (2017). Effects of biocrust on soil erosion and organic carbon losses under natural rainfall. Catena148, 117–125. doi: 10.1016/j.catena.2016.06.017
5
ChantignyM. H. (2003). Dissolved and water-extractable organic matter in soils: A review on the influence of land use and management practices. Geoderma113, 357–380. doi: 10.1016/S0016-7061(02)00370-1
6
CharambaL. V. C.HouskaT.KaiserK.KnorrK.-H.KrauseT.ChenH.et al. (2025). The deeper, the more distinct: Dissolved organic matter composition differs between soil types and diverges with depths. Front. Environ. Sci.13. doi: 10.3389/fenvs.2025.1675720
7
ChenC.ZhaoG.ZhangY.BaiY.TianP.MuX.et al. (2023). Linkages between soil erosion and long-term changes of landscape pattern in a small watershed on the Chinese Loess Plateau. Catena220, 106659. doi: 10.1016/j.catena.2022.106659
8
ChenW.GaoQ.HuH.ShaoT.ZhouC. (2024). Microbial control of soil DOM transformation during the vegetation restoration in the Loess Plateau. Plant Soil504, 385–404. doi: 10.1007/s11104-024-06627-5
9
CoryR. M.McKnightD. M. (2005). Fluorescence spectroscopy reveals ubiquitous presence of oxidized and reduced quinones in dissolved organic matter. Environ. Sci. Technol.39, 8142–8149. doi: 10.1021/es0506962
10
D'AndrilliJ.SilvermanV.BuckleyS.Rosario-OrtizF. L. (2022). Inferring ecosystem function from dissolved organic matter optical properties: A critical review. Environ. Sci. Technol.56, 11146–11161. doi: 10.1021/acs.est.2c04240
11
FanX.WangY.QuY.BaiE. (2025). Effects of vegetation restoration on soil organic carbon on the Loess Plateau, China using a combined remote sensing and process-based modeling approach. Geoderma456, 117260. doi: 10.1016/j.geoderma.2025.117260
12
FröbergM.JardineP. M.HansonP. J.SwanstonC. W.ToddD. E.TarverJ. R.et al. (2007). Low dissolved organic carbon input from fresh litter to deep mineral soils. Soil Sci. Soc Am. J.71, 347–354. doi: 10.2136/sssaj2006.0188
13
GaoJ.LiangC.ShenG.LvJ.WuH. (2017). Spectral characteristics of dissolved organic matter in various agricultural soils throughout China. Chemosphere176, 108–116. doi: 10.1016/j.chemosphere.2017.02.104
14
HanB.ChenL.XiaoK.ChenR.CaoD.YuL.et al. (2024). Characteristics of dissolved organic matter (DOM) in Chinese farmland soils under different climate zone types: A molecular perspective. J. Environ. Manage.350, 119695. doi: 10.1016/j.jenvman.2023.119695
15
HanC.LiW.LiM.GuoY.LiM.WuH. (2023). Investigating the vertical distribution of dissolved organic matter in 5-m soil profiles in farmland and typical woodland on the southern loess plateau. Environ. Res.219, 115017. doi: 10.1016/j.envres.2022.115017
16
HeN.GaoX.GuoD.WuY.GeD.ZhaoL.et al. (2024). The dimensions of deep-layer soil desiccation and its impact on xylem hydraulic conductivity in dryland tree plantations. Hydrol. Earth Syst. Sci.28, 1897–1914. doi: 10.5194/hess-28-1897-2024
17
HelmsJ. R.StubbinsA.RitchieJ. D.MinorE. C.KieberD. J.MopperK. (2008). Absorption spectral slopes and slope ratios as indicators of molecular weight, source, and photobleaching of chromophoric dissolved organic matter. Limnol. Oceanogr.53, 955–965. doi: 10.4319/lo.2008.53.3.0955
18
HensgensG.LechtenfeldO. J.GuillemetteF.LaudonH.BerggrenM. (2021). Impacts of litter decay on organic leachate composition and reactivity. Biogeochemistry154, 99–117. doi: 10.1007/s10533-021-00799-3
19
HuY.DengQ.KattererT.OlesenJ. E.YingS. C.Ochoa-HuesoR.et al. (2024). Depth-dependent responses of soil organic carbon under nitrogen deposition. Glob. Chang Biol.30, e17247. doi: 10.1111/gcb.17247
20
HuguetA.VacherL.RelexansS.SaubusseS.FroidefondJ. M.ParlantiE. (2009). Properties of fluorescent dissolved organic matter in the Gironde Estuary. Org. Geochem.40, 706–719. doi: 10.1016/j.orggeochem.2009.03.002
21
JonesD.WillettV. (2006). Experimental evaluation of methods to quantify dissolved organic nitrogen (DON) and dissolved organic carbon (DOC) in soil. Soil Biol. Biochem.38, 991–999. doi: 10.1016/j.soilbio.2005.08.012
22
KaiserK.KalbitzK. (2012). Cycling downwards – dissolved organic matter in soils. Soil Biol. Biochem.52, 29–32. doi: 10.1016/j.soilbio.2012.04.002
23
KnapenA.PoesenJ.Galindo‐MoralesP.BaetsS. D.PalsA. (2007). Effects of microbiotic crusts under cropland in temperate environments on soil erodibility during concentrated flow. Earth Surf. Processes Landforms32, 1884–1901. doi: 10.1002/esp.1504
24
KorakJ. A.DotsonA. D.SummersR. S.Rosario-OrtizF. L. (2014). Critical analysis of commonly used fluorescence metrics to characterize dissolved organic matter. Water Res.49, 327–338. doi: 10.1016/j.watres.2013.11.025
25
KothawalaD. N.KellermanA. M.CatalánN.TranvikL. J. (2021). Organic matter degradation across ecosystem boundaries: The need for a unified conceptualization. Trends Ecol. Evol.36, 113–122. doi: 10.1016/j.tree.2020.10.006
26
LavalleeJ. M.SoongJ. L.CotrufoM. F. (2020). Conceptualizing soil organic matter into particulate and mineral-associated forms to address global change in the 21st century. Glob. Chang Biol.26, 261–273. doi: 10.1111/gcb.14859
27
LeinemannT.PreusserS.MikuttaR.KalbitzK.CerliC.HöschenC.et al. (2018). Multiple exchange processes on mineral surfaces control the transport of dissolved organic matter through soil profiles. Soil Biol. Biochem.118, 79–90. doi: 10.1016/j.soilbio.2017.12.006
28
LiC.FuB.WangS.StringerL. C.WangY.LiZ.et al. (2021). Drivers and impacts of changes in China’s drylands. Nat. Rev. Earth Environ.2, 858–873. doi: 10.1038/s43017-021-00226-z
29
LiX.DaiX.TakahashiJ.LiN.JinJ.DaiL.et al. (2014). New insight into chemical changes of dissolved organic matter during anaerobic digestion of dewatered sewage sludge using EEM-PARAFAC and two-dimensional FTIR correlation spectroscopy. Bioresour. Technol.159, 412–420. doi: 10.1016/j.biortech.2014.02.085
30
LiW.JiaX.LiM.WuH. (2019). Insight into the vertical characteristics of dissolved organic matter in 5-m soil profiles under different land-use types on the Loess Plateau. Sci. Total Environ.692, 613–621. doi: 10.1016/j.scitotenv.2019.07.339
31
LiuF.ZhangG. H.SunL.WangH. (2015). Effects of biological soil crusts on soil detachment process by overland flow in the Loess Plateau of China. Earth Surf. Processes Landforms41, 875–883. doi: 10.1002/esp.3870
32
LvJ.ShiJ.WangZ.PengY.WangX. (2023). Effects of erosion and deposition on the extent and characteristics of organic carbon associated with soil minerals in Mollisol landscape. Catena228, 107190. doi: 10.1016/j.catena.2023.107190
33
MeilleurC.KamulaM.KuzykZ. A.GuéguenC. (2023). Insights into surface circulation and mixing in James Bay and Hudson Bay from dissolved organic matter optical properties. J. Mar. Syst.238, 103841. doi: 10.1016/j.jmarsys.2022.103841
34
MengY.LiP.XiaoL.HuB.ZhangC.YangS.et al. (2024). Differences in dissolved organic matter and analysis of influencing factors between plantations pure and mixed forest soils in the loess plateau. Front. For. Global Change7. doi: 10.3389/ffgc.2024.1344784
35
MessingI.ChenL.HesselR. (2003). Soil conditions in a small catchment on the Loess Plateau in China. Catena54, 45–58. doi: 10.1016/s0341-8162(03)00056-0
36
NaipalV.CiaisP.WangY.LauerwaldR.GuenetB.Van OostK. (2018). Global soil organic carbon removal by water erosion under climate change and land use change during AD 1850–2005. Biogeosciences15, 4459–4480. doi: 10.5194/bg-15-4459-2018
37
NakhavaliM.LauerwaldR.RegnierP.GuenetB.ChadburnS.FriedlingsteinP. (2020). Leaching of dissolved organic carbon from mineral soils plays a significant role in the terrestrial carbon balance. Glob. Chang Biol.27, 1083–1096. doi: 10.1111/gcb.15460
38
NasonovaA.LevyG. J.RinotO.EshelG.BorisoverM. (2022). Organic matter in aqueous soil extracts: Prediction of compositional attributes from bulk soil mid-IR spectra using partial least square regressions. Geoderma411, 115678. doi: 10.1016/j.geoderma.2021.115678
39
OhnoT. (2002). Fluorescence inner-filtering correction for determining the humification index of dissolved organic matter. Environ. Sci. Technol.36, 742–746. doi: 10.1021/es0155276
40
OhnoT.BroR. (2006). Dissolved organic matter characterization using multiway spectral decomposition of fluorescence landscapes. Soil Sci. Soc Am. J.70, 2028–2037. doi: 10.2136/sssaj2006.0005
41
PeixotoL.ElsgaardL.RasmussenJ.KuzyakovY.BanfieldC. C.DippoldM. A.et al. (2020). Decreased rhizodeposition, but increased microbial carbon stabilization with soil depth down to 3.6 m. Soil Biol. Biochem.150, 108008. doi: 10.1016/j.soilbio.2020.108008
42
PengY.XuH.WangZ.ShiJ.LvJ.WangX. (2023). Responses of the content and spectral characteristics of dissolved organic matter in intercropping soil to drought in northeast China. Plant Soil506, 471–485. doi: 10.1007/s11104-023-05931-w
43
PrăvălieR. (2016). Drylands extent and environmental issues. A global approach. Earth Sci. Rev.161, 259–278. doi: 10.1016/j.earscirev.2016.08.003
44
RenT.CaiA. (2025). Global patterns and drivers of soil dissolved organic carbon concentrations. Earth Syst. Sci. Data17, 2873–2885. doi: 10.5194/essd-17-2873-2025
45
RenT.Ukalska-JarugaA.SmreczakB.CaiA. (2024). Dissolved organic carbon in cropland soils: A global meta-analysis of management effects. Agric. Ecosyst. Environ.371, 109080. doi: 10.1016/j.agee.2024.109080
46
RinotO.BorisoverM.LevyG. J.EshelG. (2021). Fluorescence spectroscopy: A sensitive tool for identifying land-use and climatic region effects on the characteristics of water-extractable soil organic matter. Ecol. Indic.121, 107103. doi: 10.1016/j.ecolind.2020.107103
47
Riveras-MuñozN.SeitzS.WitzgallK.RodríguezV.KühnP.MuellerC. W.et al. (2022). Biocrust-linked changes in soil aggregate stability along a climatic gradient in the Chilean Coastal Range. Soil8, 717–731. doi: 10.5194/soil-8-717-2022
48
Sae-TunO.KeiblingerK. M.RosingerC.MentlerA.MayerH.BodnerG. (2023). Characterization of aggregate-stabilized dissolved organic matter release - A novel approach to determine soil health advances of conservation farming systems. Plant Soil488, 101–119. doi: 10.1007/s11104-022-05713-w
49
SchimelD. S. (2010). Climate. Drylands in the earth system. Science327, 418–419. doi: 10.1126/science.1184946
50
ShafiquzzamanM.AhmedA. T.AzamM. S.RazzakA.AskriB.HassanH. F.et al. (2014). Identification and characterization of dissolved organic matter sources in Kushiro river impacted by a wetland. Ecol. Eng.70, 459–464. doi: 10.1016/j.ecoleng.2014.06.023
51
SpeetjensN. J.TanskiG.MartinV.WagnerJ.RichterA.HugeliusG.et al. (2022). Dissolved organic matter characterization in soils and streams in a small coastal low-Arctic catchment. Biogeosciences19, 3073–3097. doi: 10.5194/bg-19-3073-2022
52
StedmonC. A.BroR. (2008). Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnology Oceanography: Methods6, 572–579. doi: 10.4319/lom.2008.6.572
53
SunZ.CuiJ.TangX.ChenL.TangJ. (2026). Temporal and spatial variability in soil dissolved organic matter under different land-use types in Southwest China. J. Soil Sci. Plant Nutr. 26, 1401–1417. doi: 10.1007/s42729-025-02938-3
54
SzyjaM.FeldeV.LuckelS.TabarelliM.LealI. R.BudelB.et al. (2023). Biological soil crusts decrease infiltration but increase erosion resistance in a human-disturbed tropical dry forest. Front. Microbiol.14, 1136322. doi: 10.3389/fmicb.2023.1136322
55
TeixeiraP. P. C.VidalA.TeixeiraA. P. M.SouzaI. F.HurtarteL. C. C.SilvaD. H. S.et al. (2024). Decoding the rhizodeposit-derived carbon’s journey into soil organic matter. Geoderma443, 116811. doi: 10.1016/j.geoderma.2024.116811
56
TianS.YaoS.ZhuS.LiP.ZhangT.SuX.et al. (2025). Evaluating soil dissolved organic matter as a proxy for soil organic matter properties across diverse ecosystems. Soil Biol. Biochem.204, 109752. doi: 10.1016/j.soilbio.2025.109752
57
van DijkA. I. J. M.KeenanR. J. (2007). Planted forests and water in perspective. For. Ecol. Manage.251, 1–9. doi: 10.1016/j.foreco.2007.06.010
58
WangY.HuH.ZhouY.ZhangB.LiS.LiuJ.et al. (2023b). Evolving characteristics of dissolved organic matter in soil profiles during 56 years of revegetation in Mu Us Sandy Land. Plant Soil497, 567–584. doi: 10.1007/s11104-023-06414-8
59
WangJ.QinX.TanY.DuY.TudiY.YangY.et al. (2024). Impact of grass cover on the soil physicochemical properties in China’s orchards: A meta-analysis. Agrofor. Syst.98, 1745–1758. doi: 10.1007/s10457-024-00985-w
60
WangH.WangJ.ZhangG. (2021). Impact of landscape positions on soil erodibility indices in typical vegetation-restored slope-gully systems on the Loess Plateau of China. Catena201, 105235. doi: 10.1016/j.catena.2021.105235
61
WangH.ZhangG.LiuF.GengR.WangL. (2017a). Temporal variations in infiltration properties of biological crusts covered soils on the Loess Plateau of China. Catena159, 115–125. doi: 10.1016/j.catena.2017.08.009
62
WangH.ZhangG.LiuF.GengR.WangL. (2017b). Effects of biological crust coverage on soil hydraulic properties for the Loess Plateau of China. Hydrol. Processes31, 3396–3406. doi: 10.1002/hyp.11263
63
WangH.ZhangQ.WangJ. (2023a). Spatial variation in soil water on a hillslope with ephemeral gullies restored by different vegetation restoration modes on the Loess Plateau. Catena224, 107001. doi: 10.1016/j.catena.2023.107001
64
WeidhunerA.HanauerA.KrauszR.CrittendenS. J.GageK.SadeghpourA. (2021). Tillage impacts on soil aggregation and aggregate-associated carbon and nitrogen after 49 years. Soil Tillage Res.208, 104878. doi: 10.1016/j.still.2020.104878
65
WeishaarJ. L.AikenG. R.BergamaschiB. A.FramM. S.FujiiR.MopperK. (2003). Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol.37, 4702–4708. doi: 10.1021/es030360x
66
WilliamsC. J.FrostP. C.XenopoulosM. A. (2013). Beyond best management practices: pelagic biogeochemical dynamics in urban stormwater ponds. Ecol. Appl.23, 1384–1395. doi: 10.1890/12-0825.1
67
WilsonH. F.XenopoulosM. A. (2009). Effects of agricultural land use on the composition of fluvial dissolved organic matter. Nat. Geosci.2, 37–41. doi: 10.1038/ngeo391
68
WitzgallK.SteinerF. A.HesseB. D.Riveras-MuñozN.RodríguezV.TeixeiraP. P. C.et al. (2024). Living and decaying roots as regulators of soil aggregation and organic matter formation—from the rhizosphere to the detritusphere. Soil Biol. Biochem.197, 109503. doi: 10.1016/j.soilbio.2024.109503
69
WuD.WangH.ZhangQ.WangJ.LiM. (2025). Dissolved organic matter characteristics of soil aggregates on slopes covered by natural successional vegetations with different root structures. Geoderma459, 117389. doi: 10.1016/j.geoderma.2025.117389
70
XuZ.WeiH.FanW.WangX.HuangB.LuN.et al. (2018). Energy modeling simulation of changes in ecosystem services before and after the implementation of a Grain-for-Green program on the Loess Plateau—A case study of the Zhifanggou valley in Ansai County, Shaanxi Province, China. Ecosyst. Serv.31, 32–43. doi: 10.1016/j.ecoser.2018.03.013
71
YairA.AlmogR.VesteM. (2011). Differential hydrological response of biological topsoil crusts along a rainfall gradient in a sandy arid area: Northern Negev desert, Israel. Catena87, 326–333. doi: 10.1016/j.catena.2011.06.015
72
YoungK. E.ReedS. C.FerrenbergS.FaistA.WinklerD. E.CortC.et al. (2021). Incorporating biogeochemistry into dryland restoration. Bioscience71, 907–917. doi: 10.1093/biosci/biab043
73
ZethofJ. H. T.CammeraatE. L. H.Nadal-RomeroE. (2019). The enhancing effect of afforestation over secondary succession on soil quality under semiarid climate conditions. Sci. Total Environ.652, 1090–1101. doi: 10.1016/j.scitotenv.2018.10.235
74
ZhangE.MengC.QuJ.ZhuZ.NiuJ.WangL.et al. (2025). Dual effects of Caragana korshinskii introduction on herbaceous vegetation in Chinese desert areas: short-term degradation and long-term recovery. Plant Soil. doi: 10.1007/s11104-025-07959-6
75
ZhangJ.WangH.WuD.ZhangQ.WangJ.LiM. (2026). Spatial heterogeneity of dissolved organic matter in vegetation-restored gully systems as shaped by coupled erosional and ecological processes. Catena262, 109637. doi: 10.1016/j.catena.2025.109637
76
ZhaoW.ZhangR.HuangC.WangB.CaoH.KoopalL. K.et al. (2016). Effect of different vegetation cover on the vertical distribution of soil organic and inorganic carbon in the Zhifanggou Watershed on the loess plateau. Catena139, 191–198. doi: 10.1016/j.catena.2016.01.003
Summary
Keywords
dissolved organic matter (DOM), land use, management intensity, optical indicators, vegetation traits
Citation
Wang Y, Zhang Q, Li N, Zhang Y, Chen G, Wu D, Gao G, Zhang Q, Zhang F, Wang J, Li M and Wang H (2026) Soil dissolved organic matter dynamics under dryland restoration: disentangling land use, vegetation, and management controls. Front. Plant Sci. 17:1934903. doi: 10.3389/fpls.2026.1934903
Received
11 July 2026
Revised
03 August 2026
Accepted
13 August 2026
Published
26 August 2026
Volume
17 - 2026
Edited by
Jianhua Xiao, Chinese Academy of Sciences (CAS), China
Reviewed by
Binhua Zhao, Xi’an University of Technology, China
Chen Fangxin, Southwest University, China
Updates
Copyright
© 2026 Wang, Zhang, Li, Zhang, Chen, Wu, Gao, Zhang, Zhang, Wang, Li and Wang.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Qing-wei Zhang, zhangqingwei@nwsuaf.edu.cn
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.