Strip thinning and filling reduces litter humus carbon loss in Cupressus funebris plantations: relative contributions of microclimate, litter substrate and microbial biomass

Alkaline-extractable litter humus carbon constitutes the precursor pool of soil organic matter (SOM) generated via litter biotransformation. However, few systematic studies have explored litter humification responses to diverse management practices in subtropical Cupressus funebris (cypress) plantations. We conducted a 1,277-day in situ litterbag decomposition experiment encompassing four stand treatments: an unmanaged pure plantation (UMP, control), artificial forest gap creation (AFG), strip thinning and filling (STF), and density-based ecological thinning (ETD). We quantified four core humification metrics across arbor, shrub, herb, and field-mass-proportioned mixed litter: alkaline-extractable humus carbon (HUc), net relative change in humus carbon (ΔHUc), humification degree (HD), and cumulative intensity of humification (HCI). The exploratory pairwise contrasts suggested that STF exhibited relatively higher HUc and HD together with less‑negative ΔHUc compared with AFG (p < 0.05); no significant differences in HUc and ΔHUc were observed between STF versus UMP, or STF versus ETD. Shrub litter exhibited markedly greater HUc and HD than arbor, herb, and mixed litter (p < 0.05). Hierarchical partitioning (HP) revealed litter substrate as the dominant factor for humification variation, accounting for 64.59 %, 71.15 %, 42.69 %, and 46.60 % of explained variance in UMP, AFG, STF, and ETD, respectively; climatic variables also exerted notable regulatory effects, particularly within STF stands. Further partial least squares path modeling (PLS‑PM) demonstrated that climatic conditions exerted strong, statistically significant direct negative effects on litter humification in all stands, even though its relative explained variance from HP was moderate. Among managed stands, STF yielded a negative point-estimate for the direct microbial-to-humification path (β = -0.422, bootstrap 95% CI -0.485 to 0.444, includes zero). By contrast, the unmanaged UMP stand had a significant positive microbial direct effect on litter humification. Significant negative direct litter‑substrate effects on humification were observed in UMP, AFG and ETD, whereas this substrate path was non-significant in STF. Overall, STF effectively mitigated litter humus carbon loss, and its humus-preserving capacity could be further enhanced by promoting understory shrub recruitment. This study provides empirical evidence for optimizing management practices to conserve litter humus in low-yield subtropical cypress plantations, and replicated multi-stand experiments are needed for broader generalization.

Forest plantation construction and ecological restoration represent core strategies to elevate forest stand quality and ecological benefits, thereby promoting litter humus retention across terrestrial ecosystems (1–3). Since the 1980s, management practices guided by forest ecological principles and community succession theory-including artificial forest gap (AFG) creation, strip thinning and filling (STF), and density-based ecological thinning (ETD)-have been extensively implemented in Cupressus funebris (cypress) plantations across the upper Yangtze River basin. Such management practices have proven effective in optimizing vertical stand structure and boosting overall stand productivity (4–7). Moreover, litter humification acts as a pivotal biogeochemical bridge linking aboveground vegetation and belowground material and energy cycling, as it controls the production of humic precursors that constitute soil organic matter (SOM) pools (8–10). Despite its indispensable role in guiding evidence-based management and the long-term sustainable development of cypress plantations, comprehensive, systematic research elucidating how humification responds to varied stand management practices is still lacking.

As core abiotic and biotic regulators governing litter humification, climatic conditions, litter substrate and microbial characteristics exert dual regulatory effects rather than consistent unidirectional influences across all decomposition stages and plantations (11, 12). As a key type of anthropogenic disturbance, forest management practices reshape understory climatic conditions (13, 14). Such modifications substantially regulate stand hydrothermal conditions, litter substrate, and microbial characteristics, ultimately exerting profound influences on litter humification (15–17). First, modulation of stand density via management practice reshapes canopy openness and surface hydrothermal conditions, yet the moisture balance varies drastically by season and openness size and type. Small, evenly distributed openness retain moderate litter moisture favorable to microbial humus synthesis, while large openness triggers intense diurnal temperature fluctuations and accelerated surface evaporation, creating drought stress that suppresses humus-forming microbial taxa such as mesophilic bacteria (18, 19). Additionally, broad-leaved and coniferous tree species differ intrinsically in canopy rainfall interception capacity (20); however, greater interception does not consistently raise soil moisture, as reduced throughfall may be offset by lower surface evaporative loss under dense canopies, ultimately leading to the variations in litter humification (9, 10, 18). Beyond climatic conditions modification, litter substrate also drives bidirectional humification responses. Moderate lignin content serve as critical recalcitrant precursors for stable humus, whereas excessively high lignin: N stoichiometric imbalance limits microbial enzyme secretion and slows organic carbon transformation (11, 21, 22). Meanwhile, moderate nutrient availability (balanced C/N, C/P ratios) stimulates decomposer proliferation to boost humus accumulation, yet surplus labile nitrogen can trigger priming effects that accelerate the mineralization of newly formed humus fractions in the short term (11, 21, 22). Anthropogenic disturbances also reshape the microbial characteristics. Appropriate stand density might increase the abundance of humus-related fungi and extracellular oxidative enzymes, whereas severe canopy disturbance disrupts stable microbial characteristics, impairing continuous humification (23–25). These mutually offsetting positive and negative pathways jointly complicate predictions of how management practices alter litter humification potentials in subtropical cypress plantations.

Forest management practices do not universally generate multi-layer understory vegetation and mixed litter inputs, such community transformation could be observable under appropriate types and intensities of management practices to low-efficiency plantations in this subtropical region (7, 26). Unmanaged closed plantations feature dense monospecific arbor canopies that block understory light, restraining shrub and herb colonization and producing litter pools dominated solely by leaf material (20, 27). Conversely, over-intensive logging eliminates stable understory microhabitats, triggering weed monopolization and homogenizing litter composition rather than forming diversified mixed litter (6, 17). Only appropriately sized openness balance light availability and soil resource supply, facilitating coexistence of arbor, shrub and herbaceous species and generating composite litter mixtures with complementary chemical traits (28, 29). Mixed litter with divergent labile and recalcitrant carbon pools improves litter layer porosity and water holding capacity, buffers extreme temperature swings during hot and dry seasons, and delivers balanced nutrient inputs to decomposers, all of which create a stable microenvironment conducive to sustained humus formation (30–32). Nevertheless, mixed litter systems also carry inherent inhibiting risks when standing biomass becomes overly abundant. Excessive litter accumulation builds thick surface organic mats that impede gas exchange and form anaerobic microsites, while uneven stoichiometric matching among coexisting litter species disrupts coordinated microbial metabolic processes (33, 34), both of which significantly retard humus transformation rates as documented by prior litter decomposition experiments. The dual facilitative and inhibitory outcomes of diversified litter inputs further highlight that the response of litter humification to management practices is contingent on vegetation structure, litter substrate and decomposition stage, and cannot be generalized as a uniformly positive ecological benefit across all plantation backgrounds.

Cypress plantations represent the dominant forest type across the hilly regions of the Sichuan Basin in China and are largely categorized as low-efficiency stands. As a critical component of the upper Yangtze River ecological shelterbelt, these cypress plantations perform vital ecological functions including soil and water conservation and regional environmental amelioration. Despite the recognized ecological value of modified cypress stands, how diverse management practices regulate litter humification remains poorly understood. To fill this research gap, we carried out a 1277-day litterbag decomposition experiment in cypress plantations under AFG, STF and ETD practices. The objectives of this study were (1): to quantify the HUc and ΔHUc in arbor, shrub, herb and mixed litter across different managed stands (2); to evaluate litter humification dynamics based on HD and HCI. Specifically, we sought to examine whether litter humification exhibits consistent response patterns to management practices alongside other well-documented ecological improvements (e.g., elevated plant biodiversity and stand productivity). The findings of this work can offer empirical and theoretical references for the sustainable management of cypress plantations and long-term ecological security protection across the upper Yangtze River basin.

Study area, experimental setup and litterfall collection

This study was conducted in Yanting County, Mianyang City, Sichuan Province (31°15′~31°18′N, 105°25′~105°29′E), located in the northern-central part of the Sichuan Basin, with an elevation ranging from 400 m to 650 m. This region is characterized by a typical hilly geomorphology, where calcareous purple soils prevail across the landscape. The average annual temperature at the sampling site is 17.3 °C, and an annual rainfall of approximately 826 mm, concentrated mainly in summer. The dominant forest type in this region is the cypress plantation, with cypress as the main arbor species. The main understory shrub species are Coriaria nepalensis and Vitex negundo, and the dominant herb species include Carex stipitinux, Cyclosorus acuminatus, Imperata cylindrica and Pogonatherum crinitum (5, 7).

From 2005 to 2008, the Sichuan Academy of Forestry implemented three forest management practices in local cypress plantations: AFG, STF and ETD. For the AFG practice, artificial logging was used to create forest gaps of 50–200 m 2. The STF practice produced 4–10 m wide strips via logging. The ETD practice implemented selective thinning with an intensity of 10%-25% of the original stand area. All logging residues were removed from experimental stands. Previous local studies confirmed that the optimal effects of each practice occurred under the following conditions: 90–100 m 2 gaps for AFG, 6 m wide strips for STF, and 20% thinning intensity for ETD. Therefore, one representative cypress plantation stand was selected for each of the three management practices (AFG, STF, ETD), with an unmanaged pure cypress plantation (UMP) as the control, all implemented from 2005–2008 with standardized optimal management intensities validated by local regional trials. Within each individual management stand, five spatially independent 20 m × 20 m replicate plots were established as subsample units for litterbag layout. To ensure plot homogeneity, all plots were at similar altitudes and slopes, and the distance between any plots was at least 50 m. We conducted a detailed vegetation survey for each plot and removed surface litter before the experiment. Detailed site information was presented in Supplementary Table 1.

Fresh leaf litter of dominant arbor (cypress), shrub (Coriaria nepalensis + Vitex negundo) and herb species (Carex stipitinux + Cyclosorus acuminatus + Imperata cylindrica + Pogonatherum crinitum) was collected from all plots in November 2020. Litter samples were transported to the laboratory and air-dried at room temperature for more than two weeks, then classified into three groups: arbor litter, shrub litter and herb litter. After drying to a constant weight, the litter was prepared by mixing it according to the natural mass ratio (arbor 1: shrub 1: herb 2) in the field. Each litter type (arbor, shrub, herb, mixture) was placed into 20 cm × 20 cm nylon litterbags, with 20 g dry litter per bag. The upper side of each litterbag had a mesh size of 1 mm × 1 mm, and the soil-contacting side had a mesh size of 0.5 mm × 0.5 mm. A total of 4400 litterbags (4 management practices × 4 litter types × 5 plots × 5 replicated litterbags × 11 sampling times) were placed on the soil surface on 30 December, 2020, with a spacing of no less than 5 cm between adjacent bags. All surface litter, roots and stones on plot topsoil were thoroughly cleared before litterbag placement to eliminate initial background impurities. In addition to initial day-0 baseline samples, litterbags were collected at 11 subsequent time points after deployment: 90, 181, 273, 365, 455, 546, 638, 729, 911, 1094 and 1277 days, yielding 12 total sampling occasions. Corresponding collection dates were 31 March, 30 June, 30 September and 31 December in 2021 and 2022; 30 June and 30 December in 2023; and the final collection on 30 June 2024. Five litterbags of each litter type were collected within each plot, yielding 400 litterbags in total for a single sampling campaign. Visible mineral soil clods, attached fine roots and extraneous fallen fresh leaves on litterbag surfaces were manually brushed off in situ immediately after collection. Additional residual soil particles, fungal mycelium and foreign litter fragments by visual inspection were further removed indoors. The five replicate litter samples of the same type from each plot were thoroughly combined into a composite sample. Each composite sample was subsequently divided into three subsamples. The first fresh subsample was preserved at 4 °C for measurements of microbial biomass carbon (MBC), microbial biomass nitrogen (MBN), and microbial biomass phosphorus (MBP). The second subsample was oven-dried at 70 °C for 48 h to constant weight to determine litter mass loss and moisture content. The third subsample was air-dried for analysing alkali-extracted humus characteristics.

Sample analyses

Total humus carbon was extracted using a mixed solution of 0.1 mol L -1 NaOH + 0.1 mol L -1 Na 4 P 2 O 7 (100 mL per sample). For each extraction, 0.5 g of crushed leaf litter was added to the mixed solution and heated at 80 °C for 30 min (9, 35). The extracts was filtered through a 0.45 μm membrane, and the alkaline-extractable humus carbon (HUc) content was determined using a total organic carbon (TOC) analyzer (vario TOC cube/vario TOC select, Elementar Analysensysteme GmbH, Hanau, Germany). In this study, all litter samples received identical extraction, digestion and measurement procedures; consistent extraction bias across all experimental groups ensures relative inter-treatment comparisons of HUc remain statistically interpretable.

To explore the effects of litter substrate and microbial biomass on litter humification, we determined decomposed litter TOC, total nitrogen (TN), total phosphorus (TP), cellulose and lignin, as well as MBC, MBN, and MBP. The determination methods were as follows: TOC by dichromate oxidation, TN by Kjeldahl (KDN, Top Ltd., Zhejiang, China), TP by phosphomolybdenum yellow spectrophotometry (TU-1800, Puxi Ltd., Beijing, China), cellulose and lignin by the Van Soest method (18, 32), and MBC, MBN and MBP by the chloroform fumigation-extraction method (36). Litter stoichiometry ratios were also considered and included in the subsequent analysis, as they are considered crucial factors governing litter humification. The temperatures of leaf litter in litterbags in the four cypress stands were obtained every two hours using iButton automatic recorders (iButton DS1923-F5, Maxim/Dallas Semiconductor, Sunnyvale, CA, USA).

Data calculations and statistical analysis

The net relative change of humus carbon (ΔHUc), humification degree (HD), cumulative intensity of humificaition (HCI) of the leaf litter were calculated were calculated using Equations 1–3, respectively:

where HUc ini and HUc t represent the initial HUc content and the HUc contents at each decomposition stage, respectively; M ini and M t are the initial litter mass and the remaining oven-dry litter mass, respectively, at each stage (t = 90, 181, 273, 365, 455, 546, 638, 729, 911, 1094, and 1277 days); HUc and TOC are the contents of total humus carbon and total organic carbon, respectively; DΔ t is the decomposition duration (days) at each stage.

To quantify the response of litter humification to different management practices relative to the unmanaged control, the natural logarithm of the response ratio (RR) was calculated to quantify the effect of management practices on humification variables (HUc, ΔHUc, HD, HCI), using Equation 4:

where H U ¯ m a n a g e m e n t and H U ¯ c o n t r o l denote the mean value of each variable in the management practices and control, respectively. To quantify management practices’ effects on ΔHUc without sign bias induced by negative relative change values, the RR of ΔHUc was calculated based on absolute residual humus carbon mass (HUc × Mt, strictly positive).

Generalized additive mixed models (GAMMs, implemented via mgcv::gamm()) were fitted via the R package “ mgcv ” to address temporal autocorrelation and non-linear decomposition dynamics, replacing the linear mixed model framework based on “ lme4 ”. Each GAMM incorporated fixed categorical effects of management practice (MP), litter type (LT), and their two-way interaction, a single global penalized thin-plate spline smooth term s(DD, k = 8) for decomposition days (DD), random intercepts for sampling site, and a continuous-time first-order autoregressive residual correlation structure corCAR1(form = ~ DD | site) to account for unequal-interval repeated measurements. Note that mgcv::gam() does not support correlation structures; the gamm() function was used for this analysis. All models were estimated via restricted maximum likelihood (REML). Shapiro-Wilk tests of model residuals showed raw HUc and ΔHUc complied with normality assumptions (p > 0.05), while untransformed HD and HCI (0-100% bounded percentages) exhibited significant residual skewness. A small epsilon offset (1e-4) was applied prior to logit transformation of HD and HCI raw values to avoid infinite logit outputs at boundary values of 0 or 100%. Global significance of categorical fixed factors (management practices, litter type, management practices × litter type) was evaluated using Type-III Wald χ² tests; approximate F-tests were used to assess the significance of the decomposition smooth term s(dd). Kenward-Roger denominator degrees of freedom cannot be computed for GAMMs incorporating the corCAR1 autocorrelation structure, so conventional F-statistics with separate numerator/denominator df are unavailable. Adjusted R² and deviance explained values were reported to quantify overall model explanatory power. Marginal pairwise contrasts for management practice and litter type, averaging across the other categorical factor, were computed from fitted GAMM objects. Multiplicity correction for pairwise comparisons was applied using the false discovery rate (FDR) method. In this analysis, t = 0 (day-0) observations were excluded for ΔHUc and HCI analysis, resulting in effective sample sizes n = 880 for ΔHUc and HCI, while HUc and HD used n = 960 (including day-0 samples). Post-hoc pairwise comparisons were performed using estimated marginal means (EMM) with Tukey adjustment. Compact letter display (CLD) was used to denote significant differences among treatments. Moreover, monthly temperature (MT) and decomposition day (DD) are inherently seasonally confounded across the time-series sampling, associations for MT should therefore be interpreted as correlative.

Hierarchical partitioning (HP) analysis was conducted to compare the relative importance proportion of climatic conditions, litter substrate, and microbial biomass characteristics associated with litter humification within each stand, using the R package “ rdacca.hp ” (37, 38). Prior to finalizing the optimal HP model for each stand, we systematically conducted VIF multicollinearity screening for all candidate predictors. We sequentially removed highly collinear predictors and retained only one set of variables to eliminate deterministic dependence between components and ratios. All variables retained in the final analysis have VIF < 5, satisfying standard multicollinearity thresholds for ecological multivariate modeling, meanwhile, based on the R 2, eliminate and add relevant variables. Partial least squares path modeling (PLS-PM) was applied to quantify the complete causal chain of how forest management reshapes climatic conditions, litter substrate, and microbial biomass characteristics to alter litter humification, and test cross-stand differences in path coefficients, using the “inner plot” function in the R package “ plspm ” (39). Arrows and the path coefficients represent the direction and strengthen of the relationships between latent variables, and the model fit was evaluated using goodness of fit (GoF) statistic. For PLS-PM outputs, we use bootstrap-generated 95 % confidence intervals to judge statistical significance of standardized path coefficients, instead of relying solely on built-in package p -values. All statistical analyses were performed in R-4.5.3 (R Development Core Team, 2026), and the results were visualized using “ ggplot ” package in R-4.5.3.

Each management practice is represented by a single stand. The five plots inside each stand function as spatial subsamples nested within that stand and do not constitute independent stand-level biological replicates. Consequently, all pairwise contrasts among management practices reported in this study are exploratory descriptive observations restricted to our four study stands and should not be interpreted as universally valid treatment effects applicable to other cypress plantations.

Effects of management practices and litter type on litter humification

Management practice had a statistically significant global effect on ΔHUc (p = 0.011), whereas global main-effect tests for HUc (p = 0.053) was non-significant (Table 1). STF had relatively higher average HUc and less-negative ΔHUc compared with AFG (Table 2). No comparable exploratory differences in HUc and ΔHUc were observed between STF versus UMP or STF versus ETD (Table 2). Litter type showed significant associations with HUc and ΔHUc (p < 0.01; Table 1). Shrub litter had the highest HUc, and herb litter exhibited the greatest ΔHUc (p < 0.05; Table 2). Relative to UMP, STF showed a positive ln-response-ratio shift of (0.035 ± 0.026) for HUc, with the largest divergence for shrub litter (0.112 ± 0.04) (Figure 1a). For ΔHUc, STF yielded a mean ln-response-ratio of (-0.010 ± 0.052), and herb litter had the largest value (0.065 ± 0.052) (Figure 1b).

Global tests indicated non-significant stand-wide management-practice effects for HD (p = 0.252) and HCI (p = 0.279) (p > 0.05; Table 1). STF had relatively higher average HD and HCI compared with AFG (Table 2); given the non-significant global main-effect for HD and HCI, this pairwise finding is treated as observational rather than definitive evidence of management-driven differences. On average, STF displayed a positive correlative shift of (0.099 ± 0.036) in HD ln (response ratio) relative to AFG (p < 0.05; Figure 1c); no significant differences were detected between STF and UMP or ETD (Table 2). However, AFG exhibited a relative decrease of (−0.075 ± 0.024) in the ln(response ratio) of HCI on average compared with STF (Figure 1d). Litter type exerted a significant global effect on HD and HCI (p < 0.05; Table 1). Shrub litter had consistently elevated HD and HCI relative to other litter types (Table 2).

Temporal variations of litter humification in different cypress plantations

The global smooth term for decomposition day s(DD) was highly significant for HUc and ΔHUc (p < 0.001), indicating overall pronounced temporal dynamics in humification across the full dataset. Our GAMM uses one global smooth of decomposition day and does not include stand-specific smooth terms, so formal statistical tests for differing temporal trajectories among stands were not performed. HUc appears to follow a visual polynomial-like pattern (decline then rise) in AFG, STF and ETD, while ΔHUc shows consistent declining visual trends (Figures 2a, b). Graphically, HUc and ΔHUc values for STF and ETD gradually become higher relative to AFG and UMP as decomposition proceeds (Figures 2a, b).

The global smooth term for decomposition day s(DD) was highly significant for HD and HCI (p < 0.001), demonstrating overall strong temporal shifts in litter humification across the full dataset (Figures 3a, b). Visually, litter HD increased with increasing litter mass loss, especially in STF and ETD stands. After 1277 days of decomposition, STF stand had the highest litter HD among all stands (Figure 3a). The smooth term for decomposition day was highly significant for both HD and HCI (p < 0.001), revealing strong temporal humification patterns. HCI (HD divided by decomposition days) mechanically declines over time as HD accumulates and decomposition duration lengthens (Figure 3b).

Driving pathways of litter humification in managed cypress plantations

HP analysis indicated that climatic conditions, litter substrate and microbial biomass characteristics together explained 44.2%, 47.8%, 37.6% and 37.3% of the variations in the integrated litter humification index in the UMP, AFG, STF and ETD stands, respectively (Figure 4a). Litter substrate constituted the dominant driver in all stands, contributing 64.59% (UMP), 71.15% (AFG), 42.69% (STF), and 46.60% (ETD) of the explained variation (Figure 4a). However, among climatic conditions, moisture conditions (UMP: 15.73%, AFG: 14.71%, ETD: 13.98%) and monthly temperature (MT) (STF: 22.48%) were the key predictors in all stands (Figure 4b). For litter substrate variables, TOC (UMP: 6.96%, AFG: 19.16%, STF: 14.83%, ETD: 9.17%), cellulose (UMP: 15.99%, AFG: 23.79%, STF: 7.09%) and TP (UMP: 7.00%, AFG: 10.26%, STF: 13.67%, ETD: 5.98%) were the main contributors to humification variations (Figure 4b). In terms of microbial biomass characteristics, the MBC/MBP ratio was the key microbial factor in STF and ETD stands, accounting for 13.47% and 19.57% of variation, respectively (Figure 4b).

PLS-PM path outputs revealed that climatic conditions exerted strong significant direct negative effects on litter humification in every stand, alongside indirect regulation via modification of litter substrate and microbial biomass characteristics (Figure 5). The unmanaged UMP stand had a significant positive microbial effect (β = 0.360, p < 0.001) and significant negative substrate effect (β = -0.094, p < 0.05) (Figure 5a). In the STF stand, the microbial-to-humification path produced a negative point estimate (β = -0.422), but its bootstrap 95 % confidence interval (-0.485 to 0.444) spanned zero, so this path was not statistically significant, and the litter-substrate direct effect remained non-significant for STF (β = 0.006, p > 0.05). However, litter substrate exerted significant negative effects in AFG (β = -0.287, p < 0.001) and ETD (β = -0.278, p < 0.001) (Figures 5b–d). All reported β values represent standardized direct path effects derived from PLS-PM bootstrap outputs, rather than total effects. Manifest indicators for the latent humification construct were iteratively pruned based on HP results and model goodness-of-fit (GoF) to improve model performance for each stand. This procedure produced divergent manifest-indicator sets across the four stands. Multi-group invariance testing was therefore not applicable, and interpretations focus only on the direction and significance of path coefficients within each stand, without quantitative comparison of β magnitudes across management groups.

Effects of management practices on litter humification in cypress plantations

Litter humification is a critical process in SOM formation and greatly determines the reduction of litter humus carbon loss and facilitate and the formation of humus precursors for SOM in plantation ecosystems (3, 40). Based on widely replicated forest biogeochemical experiments reported in prior literature, standardized management practices can reshape canopy structure, understorey climatic conditions and understory vegetation, and such management-mediated shifts in abiotic and litter substrate further regulate litter humification. Although our experiment only documents correlative differences in humification indexes across four separate plantations, this established mechanistic mediation pathway provides a plausible interpretive framework for the patterns observed in our 1277-day litter incubation dataset. In our study, STF exhibited relatively higher HUc, ΔHUc, HD and HCI compared with AFG; although non-significant relative to UMP and ETD, STF still showed numerically more favorable humus-retention performance among the four stands, which partially addresses our scientific questions. The relatively lower average of HUc, ΔHUc, HD and HCI, with higher litter mass loss in the AFG stand relative to other stands, which aligns with previous findings (17, 41, 42). Based on stand-specific exploratory observations within our experiment, AFG appeared to favor accelerated material cycling, whereas STF showed a tendency toward reduced litter humus carbon loss and greater humus-precursor formation. Nevertheless, these patterns were observed in one stand per management type, and multi-stand replicated trials are required to verify whether such performance differences hold for broader cypress plantation management applications. In general, forest management practices consistently improve visible ecological benefits (e.g., biodiversity conservation and forest productivity), but their effects on cryptic biogeochemical processes such as humification are variable, which is closely related to specific management practices and implementation intensities.

Consistent with previous studies, litter humification in cypress plantations is jointly regulated by stand climatic conditions, litter substrate, and litter-associated microbial biomass characteristics (13, 14). As shown by HP results, litter substrate contributed the largest proportion of explained variance in humification variation across stands. Nevertheless, PLS-PM outputs indicated that climatic conditions exerted strong significant direct negative effects on litter humification in all stands; climate also acted as an indirect moderator by reshaping litter substrate and microbial biomass. It should be noted that HP assesses total explanatory importance (aggregating direct and indirect effects), while PLS-PM estimates direct causal path coefficients. Thus, a relatively lower proportion of variance explained for climate in HP does not contradict the existence of large-magnitude significant direct causal paths in PLS-PM. Distinct direct-effect patterns for litter substrate and microbial biomass on litter humification were observed across stands. A significant negative direct substrate effect on humification occurred in UMP (β=-0.094, p < 0.05), AFG (β=-0.287, p < 0.001), and ETD (β=-0.278, p < 0.001). By contrast, the direct substrate-to-humification path was non-significant in STF.

The unmanaged UMP stand displayed a statistically significant positive direct microbial effect on litter humification (β = 0.360, p < 0.001). For managed stands, STF yielded a negative point-estimate for the microbial-to-humification path, yet bootstrap confidence intervals included zero and thus this effect could not be confirmed as statistically significant. We therefore cannot draw firm conclusions about consistent stand-level differences in microbial direct-effect directions between managed and unmanaged plantations from our PLS-PM outputs. Observed differences in humus-retention performance across stands likely arise from combined indirect influences of stand microclimate, litter substrate and microbial properties, rather than from a single direct microbial-driven causal path. Moreover, even without a significant direct substrate-humification path, STF still had the numerically and relatively higher HUc and HD values among stands (Table 2), consistent with our core observation that STF could partly mitigate litter humus carbon loss relative to AFG. This suggests STF’s better humus-preserving performance cannot be explained by a single direct substrate-driven causal path; it also likely emerges from combined indirect influences of stand microclimate, litter substrate and microbial properties (9, 11, 13, 22). In UMP’s shaded, low-nutrient understory environment, slow microbial metabolic rates might limit the breakdown of newly formed humus, generating the positive microbial-humification association captured in the path model (18). Hierarchical partitioning results aligned with PLS-PM structural patterns: litter substrate remained the dominant explanatory variable for litter humification variation in all stands, while monthly temperature became a key climatic driver exclusively within STF stand. The numerically higher HUc ΔHUc, HD and HCI values under the STF stand (Table 2), indicating that this management practices might mitigate humus carbon loss. The key point that needs to be noted here is that the general net loss of humus carbon pool (negative ΔHUc) in all stands, meaning none of the management practices generated absolute net humus carbon accumulation over the full study period. This phenomenon arises from dual processes: newly formed humus substances from litter transformation are simultaneously subject to photodegradation, leaching and microbial mineralization, and the decomposition loss exceeds humus formation gain across all stages. Therefore, STF’s superiority lies in its ability to suppress humus carbon decomposition loss by optimizing stand microclimate, rather than accelerating humus formation to offset carbon mineralization completely.

Furthermore, numerous prior studies have validated that specialized microbial communities are central to humus transformation and formation processes. Specifically, lignin-degrading fungi dominate oxidative humus turnover, whereas bacterial taxa including Anaerolineae, Methylibium, and Gallionella can suppress organic carbon mineralization and facilitate humus accumulation (25, 43). Given that the current experiment only quantified total microbial biomass without resolving fine-scale community composition, we hypothesize that divergent microbial taxonomic assemblages induced by different management practices support the stand-specific humification disparities observed in this subtropical cypress plantation. To further clarify such underlying microbial mechanisms and enhance predictive capacity for litter humification dynamics in cypress plantations, future research should integrate high-throughput sequencing to characterize fungal and bacterial community structure, and link taxonomic functional traits to long-term humus transformation patterns.

Effects of litter type on litter humification in cypress plantations

Multiple well-documented conceptual frameworks, including nutrient transfer theory, specific compound interaction theory and trophic complementarity theory, are widely applied to interpret non-additive mixing effects during litter decomposition (44–46). It should be noted that these theories do not predict a universal positive correlation between litter mixing and decomposition rate. Instead, mixed-litter outcomes (synergistic, antagonistic, or additive) are highly context-dependent, modulated by plant species composition, mixing proportion, inherent litter substrate, decomposer community assembly, and ambient climatic conditions (32, 47). Furthermore, litter mass loss rate cannot be directly equated with humification performance, since mineralization, humus formation and humus degradation proceed simultaneously throughout decomposition, generating divergent responses of decomposition and humification to identical litter mixtures (32, 47).

In our experiment, the mixed litter (arbor: shrub: herb = 1:1:2 by natural mass ratio) exhibited lower cumulative mass loss relative to single shrub litter, while shrub litter displayed significantly higher HUc and HD values than other litter types. Direct comparison between mixed litter and shrub litter alone cannot reliably demonstrate antagonistic mixing effects or contradict existing decomposition theories, because quantitative evaluation of litter mixing non-additivity requires calculating the additive expected value of mixed litter based on the individual performance of each component weighted by their actual mixing proportion. The observed higher humification level of shrub litter also does not prove that mixed litter experienced suppressed humification relative to theoretical expectations without this standardized calculation. We offer tentative contextual explanations for the divergent humification patterns observed in this subtropical cypress plantation rather than that established theories are inapplicable or that unique humification pathways exist in our system. First, the distinct litter substrate of local dominant shrub species (lower lignin, lignin: N and C/N ratios compared to arbor and herb litter) create unique nutrient supply regimes for decomposers, which alters humus transformation efficiency independent of classical litter mixing mechanisms (32, 46, 48, 49). Second, the fixed field-derived mixing ratio used in this study generates substrate stoichiometry and physical litter layer structures different from those reported in most temperate or alpine mixed-litter studies, leading to context-specific humification responses. These two factors jointly drive the divergent humification signatures between single shrub litter and composite mixed litter under our experimental conditions. Future quantification of non-additive mixing effects (calculated observed vs. predicted HUc, HD and mass loss values), combined with measured litter chemical traits and microbial biomass indicators, would help disentangle the underlying mechanisms regulating mixed-litter humification in subtropical cypress plantations.

Temporal dynamics of litter humification in cypress plantations

Consistent with previous studies, litter humification showed distinct stage-dependent temporal dynamics, jointly driven by climatic conditions, litter substrate and microbial biomass characteristics in this study. During 0–365 d of decomposition, climatic conditions dominantly increased humus loss due to direct physical impacts on litter substrate (especially carbon and nutrient leaching) and indirect microbial mediation through enhanced substrate availability (Supplementary Figure 2) (47, 50). Different from traditional decomposition models (32, 49, 51, 52), lignin and cellulose exerted stronger control over humification in this stage than the decomposition stage of 911–1277 d, which is consistent with the finding that lignin degradation mainly occurs in the stage of 0–365 d (52). This pattern is closely linked to labile carbon quality, nutrient availability and microbial properties (21, 23), and our HP analysis also confirmed these significant correlations. As decomposition proceeded, labile litter components were gradually depleted, and the influence of climate weakened, accompanied by reduced microbial activity that dominates litter humification (9, 23). Consequently, humus retention continued to decline during 455–729 d period, which was mainly governed by residual carbon and nutrient in litter (Supplementary Figure 2). However, during the 729–1277 d period, humus retention showed an increasing trend, which is likely due to unstable humus fractions were depleted by long-term climatic and microbial effects in the stage of 0–365 d (18), and new humus began to form under the regulation of residual nutrients and humus-promoting microorganisms (e.g., Gallionella) (25, 32, 43, 48). We modelled HD as a function of cumulative litter mass loss using linear regression to compare humification accumulation trajectories among stand and litter treatments (Figure 3a), which avoids the tautological trend of HCI and captures biologically meaningful differences in long-term humus formation capacity. Although HCI showed uniform decreasing dynamics for all litter types and management practices, no inter-group significant differences were observed, indicating consistent time-normalized humification speed across our experimental gradients. A comprehensive analysis showed that although the regulatory effects of climatic conditions, litter substrate, and microbial characteristics on the HD and HCI exhibited stage-specific dynamics, their directional influences remained consistent, continuously promoting or inhibiting the humification process throughout the entire study period.

Descriptively, Figure 2 shows apparent visual differences in humification trajectories across the four plantation stands, although our GAMM with a single global smooth term does not provide formal statistical evidence for stand-specific temporal curve differences. Visually, AFG and UMP tended to show relatively higher HUc and ΔHUc values during the early decomposition phase (0-181 d). This pattern was likely due to higher temperature in AFG activated humus-related microorganisms (17, 18), while UMP stand had higher initial HUc (substrate legacy effect) (45, 49). After this stage, unstable humus fractions in AFG and UMP stands were gradually degraded by climate and leaching, leading to a sharp drop in humus retention (18). In contrast, STF and ETD stands showed a lagged but sustained increase in humification during the 365–911 d period, and their advantages became more prominent over time. At this stage, climate effects faded, and litter nutrients and cellulose became the main controlling factors (9). These findings demonstrate that forest management practices, particularly STF and ETD cypress plantations, could reduce humus carbon loss in cypress plantations through enhanced biochemical stabilization pathways, with the STF stand showing the most superior retainment effect. However, this conclusion is drawn based on visual dynamic variations, and a more accurate conclusion derived from statistical analysis requires further research.

In this 1277-day litterbag experiment conducted across four stands (one stand per management practice, with plots as nested spatial subsamples), exploratory pairwise contrasts within our study indicated that STF exhibited relatively higher HUc and HD and less-negative ΔHUc compared with AFG (p < 0.05). No corresponding differences were observed between STF and UMP or ETD. These stand-limited observations suggest a tendency for STF to reduce litter humus carbon loss relative to AFG in the present experimental setting, though this finding cannot be generalized due to the absence of independent stand-level replication. HP analysis identified litter substrate as the main source of explained variance for litter humification, while PLS-PM further revealed that climatic conditions exerted strong direct negative effects on humification alongside indirect regulatory pathways mediated by litter substrate and microbial biomass; cellulose, lignin, and elemental stoichiometry constituted the core substrate regulators. Only the unmanaged UMP stand returned a statistically-supported significant direct microbial effect; for managed stands, microbial path point-estimates varied by stand but several bootstrap confidence intervals crossed zero, limiting robust cross-stand inference for direct microbial effects. Furthermore, litter type significantly altered HUc and HD but produced no detectable shifts in HCI. Shrub litter possessed the strongest humification potential, yet simple pairwise comparison between mixed and shrub litter cannot identify additive, synergistic, or antagonistic non-additive interactions without mass-weighted predicted indices; inter-litter-type disparities represent context-specific characteristics of local subtropical cypress plantation ecosystems and do not contradict established decomposition theories. Over the study period, litter humification exhibited clear stage-dependent trajectories: continuous net humus carbon loss occurred during early and middle decomposition stages. STF’s humus-preserving advantages emerged gradually after the mid-decomposition stage, whereas warm, humid microhabitats within AFG canopy gaps accelerated both litter breakdown and humus degradation, triggering severe net humus depletion. Collectively, our stand-specific exploratory observations suggest that STF may represent a promising management option for conserving litter humus in low-yield cypress plantations, and promoting natural understory shrub recruitment could potentially further improve humus-retention performance. However, multi-stand replicated field experiments are essential to confirm these patterns before broad-scale management recommendation. Two key experimental limitations remain: we did not calculate mass-weighted metrics to quantify non-additive mixed-litter effects, and humus quantification relied exclusively on alkaline extraction. Follow-up research should integrate solid-state ¹³C nuclear magnetic resonance (NMR) and mineral-associated organic matter fractionation to characterize stable organic carbon fractions, alongside weighted litter indices to disentangle interactive decomposition dynamics of multi-species litter mixtures.