Incorporated plants were applied as aboveground biomass: radish (Raphanus sativus L.), safflower (Carthamus tinctorius L.), and white mustard (Sinapis alba L.) were harvested from fields near Nová Ves, Brno (49°06′25.7″ N, 16°18′15.3″ E) after 77 days (11 weeks; biomass A) or 87 days (12.5 weeks; biomass B; Table 1). Plants were cut at ground level, oven-dried at 60 °C to constant weight, analysed (details in the section Plant and soil analyses) and stored.

The experiment was conducted at Agricultural Research, Ltd. (Troubsko, Czech Republic; 49°17′49.8″ N, 16°49′13.0″ E). Arable topsoil (0–15 cm) was collected from fields managed conventionally in a cereal–rapeseed–pea rotation. The site is moderately warm and dry (mean annual temperature 9.0 °C, precipitation 526 mm). Geologically, it is underlain by loess/loess loam of the Bohemian Massif. The soil was classified as a Haplic Luvisol (WRB) with silty clay loam texture (USDA). The primery chemical properties were: total C 14.0 g·kg⁻¹, total N 1.60 g·kg⁻¹, P 0.10 g·kg⁻¹, S 0.15 g·kg⁻¹, Ca 3.26 g·kg⁻¹, Mg 0.24 g·kg⁻¹, K 0.23 g·kg⁻¹; pH (CaCl₂) 7.3. Soil was homogenized and sieved (2 mm, Retsch GmbH, Germany).

The trial was conducted in rectangular boxes (57 × 37 cm; 10 kg soil each). Two semipermeable polyamide mesh bags (Uhelon 130 T, 42 nm pore size, SILK and PROGRESS, Czech Republic), each containing 20 g of dry biomass (equivalent to 2 g·kg⁻¹ soil), were placed in each box and covered with 2 cm soil layer (Supplementary Figure S1). All treatments (Table 2) were arranged in triplicate (n = 3).

The experiment was conducted from 16 February to 16 June 2022 (120 days). Environmental conditions were maintained at 9/16 °C (night/day), and each box was watered weekly with 1.5 L of distilled water to simulate alternating dry and moist periods. Urea was applied at a rate of 3.44 g per box on days 15 and 35 after incubation began. This meant than soil was fertilized with urea (46% N, AGRO CS a.s., Czech Republic) at dose of 0.688 g N·kg⁻¹ (150 kg N·ha⁻¹), applied in two splits.

During sampling at the end of the trial, soil was gently removed, mesh bags collected, and soil sampled from the 0–2 mm layer beneath each bag. Residual biomass (RB; g·bag⁻¹) was oven-dried (65 °C) and weighed.

All data (plant biomass and quality traits, soil chemical and biological parameters) were analysed in R v4.3.1 (R_Core_Team, 2023). Pearson’s correlations were used to characterise relationships, interpreted as moderate (0.5–0.7), high (0.7–0.9), or very high (>0.9) (Hinkle et al., 2003). Figures were created using ggplot2 0.5–0.7 = moderate, 0.7–0.9 = high, and >0.9 = very high correlation. Figures were created using ggplot2 package (Wickham, 2016).

One-way and two-way ANOVA with Type I sums of squares were applied at α = 0.05 (Zar, 1984), using the FactoMineR (Lê et al., 2008) and Factoextra (Kassambara and Mundt, 2020) packages. Tukey’s HSD (α = 0.05) was used for post hoc comparisons. Results are presented as standard error of the mean (±SEM) (Faraway, 2014).

Model assumptions were tested at α = 0.05. Normality of residuals was evaluated with Kolmogorov–Smirnov and Anderson–Darling tests (nortest package). Homoscedasticity was assessed by Bartlett’s and Fligner–Killeen tests. Diagnostic plots included residuals vs. fitted values, Q–Q plots, Cook’s distance, and leverage plots (Faraway, 2014; Pekár and Brabec, 2016). Distributional shape was further tested using D’Agostino’s K-squared (skewness) and Anscombe–Glynn (kurtosis) tests (moments package (Komsta and Novomestky, 2022).

Chemical and nutritional traits of plant biomass, which determine decomposition rates, varied significantly with both crop species and harvest time (one-way ANOVA; partial eta-squared; Supplementary Table S1; Figures 1A–H).

Carbon content (C, % DM) was lower in radish compared to safflower and white mustard, and also differed between RadA and RadB (Figure 1A). Nitrogen content (N, % DM) did not vary among crop species but was higher in RadB than in RadA (Figure 1B). Ash content (% DM) was highest in radish for both harvests; however, later harvest increased Ash in both radish and safflower (Figure 1C).

Carbohydrates (CH, % DM) in biomass A showed a decreasing trend from radish (highest) to white mustard (lowest), whereas in biomass B radish and white mustard were comparable, and both SafB and WmuB increased with ripening time (Figure 1D). Fat (% DM) content showed the strongest contrast between harvest stages in white mustard, with the lowest values in WmuA and the highest in WmuB (Figure 1E).

Neutral detergent fibre (NDF) and acid detergent fibre (ADF, % DM) decreased in radish and safflower biomass B compared to biomass A. In contrast, WmuB had significantly higher NDF, ADF, and acid detergent lignin (ADL, % DM) than the other crops at the later harvest (Figures 1F–H). ADL in biomass A was highest in SafA but declined in SafB, while RadB and WmuB were higher than RadA and WmuA (Figure 1H).

Similar to NDF, ADF, and ADL, crude fibre (CF) and cellulose (CEL, % DM) values were highest in WmuB (Figures 1I,J). In biomass A, CEL was reduced in SafA. For safflower, CF and CEL decreased from SafA to SafB, while hemicellulose (HCEL, % DM) showed the opposite trend: highest in SafA but reduced in SafB compared to the other crops (Figure 1K).

The ratio of ADL to nitrogenous compounds (ADL_NC), an indicator of digestibility, closely followed ADL patterns. Values increased from biomass A to biomass B in radish and white mustard, but decreased in safflower (SafB < SafA; Figure 1L).

Residual biomass (RB, % DM) was significantly higher in SafA than in the other two crops, indicating slower decomposition of safflower. However, RB values were comparable across crops in biomass B (Figure 1M).

Soil pH did not differ among variants with various incorporated crops, neither in earlier (biomass A) or later maturity stage (biomass B). However, biomass B amendments generally lowered pH compared with unamended soil (Figure 2C). In particular, RadB- and WmuB-amended soils had lower pH than RadA and WmuA, respectively (Figure 2A).

Soil contents of oxidizable organic carbon (Cox), total carbon (TC), total nitrogen (TN), and their ratio (Cox/TN) were unchanged across crops within both maturity stages (Figures 3A–I). By contrast, total phosphorus (P) and potassium (K) increased in soils amended with biomass A, showing a trend from lower values in RadA to higher values in WmuA, all exceeding the unamended control (Figures 4B,H). Biomass B amendments also elevated P, K, and Mg contents relative to the control, except for P in SafB and Mg in RadB (Figures 4C,F,I). Within each maturity stage, Mg did not differ among crops (Figure 4D). P and K were highest in WmuA soils, while P was highest in WmuB soils, compared with RadA (P, K) and SafB (P) (Figures 4A,G).

Enzyme activities responded variably to biomass incorporation. β-glucosidase (GLU) activity increased in RadA- and SafA-amended soils compared with the control (Figure 5B), but showed no differences among crop treatments within each maturity stage (Figure 5A). Phosphatase (Phos) activity was higher in RadA and SafB than in the control (Figures 5E,F). Across maturity stages, biomass A amendments generally enhanced GLU (all crops) and Phos (radish and white mustard) compared to biomass B (Figure 5A).

Urease (Ure) activity increased in SafA but decreased in RadB compared with the control (Figures 6B,C). Differences among crops were also evident: WmuA and RadB had lower Ure activity than the other two crops at the same maturity stage (Figure 6A). N-acetyl-β-D-glucosaminidase (NAG) and arylsulfatase (ARS) increased in RadA relative to the control (Figures 6E,H). Biomass B amendments further raised NAG in WmuB and ARS in RadB and SafB (Figures 6F,I). Among crop comparisons, WmuA decreased ARS relative to RadA, whereas WmuB increased NAG compared to RadB and SafB (Figures 6D,G).

Enzymatic stoichiometry also shifted with crop type and maturity. The soil carbon acquisition ratio (Cacq.) and vector length (C limitation index) remained unchanged with biomass A but increased under RadB compared with other crops (Figures 7C,I). Within biomass A, WmuA showed higher Cacq. and vector length than SafA, while RadB had the highest values among biomass B treatments (Figures 7A,G).

Nitrogen acquisition ratio (Nacq.) and vector angle (N limitation index) showed opposite patterns. RadB decreased Nacq. but increased vector angle relative to the control (Figures 7F,L). Within biomass A, WmuA had lower Nacq. and higher angle than SafA. Conversely, WmuB had the highest Nacq. and lowest angle among biomass B treatments. Across crops, Nacq. declined while vector angle increased with longer maturation (biomass A vs. B; Figures 7D,J).

Pearson’s correlation analysis (Supplementary Figure S2A revealed several significant relationships. Plant C content correlated negatively with Ash content (p ≤ 0.001, r = −0.89) and with the soil Vector index (p ≤ 0.01, r = −0.65), but positively with soil urease (Ure) activity and N acquisition ratio (Nacq.) (p ≤ 0.001, r = 0.74 and 0.71, respectively). Conversely, Ash content correlated negatively with Ure and Nacq. (p ≤ 0.001, r = −0.80 and −0.73). Acid detergent fibre (ADF) correlated positively with Nacq. (p ≤ 0.01, r = 0.63), while neutral detergent fibre (NDF) and residual biomass (RB) correlated positively with soil pH (p ≤ 0.01, r = 0.61; p ≤ 0.001, r = 0.72). Acid detergent lignin (ADL) correlated negatively with β-glucosidase (GLU; p ≤ 0.01, r = −0.60). Residual biomass correlated positively with hemicellulose (HCEL; p ≤ 0.001, r = 0.83). Among soil parameters, phosphatase (Phos) correlated negatively with P, K, and Mg contents (p ≤ 0.01, r = −0.65, −0.70, −0.73).

When analysed separately for biomass A (Supplementary Figure S2B), RB correlated positively with NDF, HCEL, and ADL (p ≤ 0.001, r = 0.87, 0.93, 0.85), and negatively with ADF, cellulose (CEL), and Ash (p ≤ 0.001, r = −0.81, −0.88, −0.95).

The results of residual biomass quantification, soil trait and plant quality determination indicated that both crop choice and, more decisively, harvest timing are critical determinants of biomass quality and subsequent soil responses. Early-harvested biomass provides more labile substrates that promote larger decomposition, whereas later-harvested biomass, particularly of white mustard, is more resistant and may contribute to long-term soil organic matter stabilization. Specifically, residual biomass (RB; Figure 1M) indicated more extensive biodegradation of all crops (77%–87% mass loss), except safflower biomass A (SafA), which decomposed less (62% loss). This lower degradability of SafA was associated with higher NDF, ADL, hemicellulose, and ADL:NC ratios, and reduced cellulose and Ash. High lignin content (Jensen et al., 2005), hemicellulose (Henriksen and Breland, 1999), and elevated ADL:nitrogen ratios (Moore et al., 1999) are generally linked with structural stability and decomposition resistance. Conversely, high protein, fat, and soluble carbohydrates support easier degradation (Heal et al., 1997; Aditya and Wolfram, 2020).

Interestingly, the most degraded biomasses such as RadA and WmuB did not show higher nitrogenous compound contents, challenging the assumption that N-rich biomass always decomposes easily. Instead, both were rich in fat, fibre, and cellulose. Although these parameters cause moderately recalcitrant decay (Henriksen and Breland, 1999; Zoghlami and Paës, 2019), cruciferous crops (radish, white mustard) tended to degrade more easily, consistent with previous observations (Kucerik et al., 2024). By contrast, safflower decomposed less easily, despite relatively low ADL and fibre fractions in SafB, highlighting inconsistency with previous reports of safflower green manure contributing substantially to soil C (Kucerik et al., 2024).

As aforementioned, decomposition was shaped by both crop type and maturity stage (two-way ANOVA, Supplementary Table S1). In some traits (e.g., fat, NDF, hemicellulose), harvest stage had a stronger effect than crop identity. Correlation analysis confirmed that RB was positively associated with hemicellulose and NDF, and negatively with lignin, cellulose, and Ash (Supplementary Figure S2). This suggests that decomposition depends not only on the abundance of resistant polymers, but also on the specific balance between moderately degradable (cellulose, hemicellulose) and resistant (lignin, Ash) compounds.

These results indicate a fundamental difference between crucifers and safflower: cruciferous residues appear more degradable despite moderately recalcitrant fibre fractions, possibly due to their biochemical composition (e.g., glucosinolates and associated breakdown products), while safflower shows unexpectedly slow decomposition even at lower ADL contents. Such a decoupling between bulk ADL and decay rate is consistent with evidence that lignin chemistry and phenolic cross-linking (e.g., ferulate-mediated linkages) can physically and chemically protect labile carbohydrates, thereby constraining decomposition independent of ADL concentration (Hatfield et al., 2017; Talbot et al., 2012).

Therefore, the results imply that cruciferous crops appear more degradable, making them better suited for short-term nutrient release, while safflower may stabilize soil C. A combination of cruciferous residues with safflower could therefore balance rapid nutrient turnover with longer-term carbon sequestration, a strategy that could be agronomically exploited for soil fertility management.

Previous studies often report that green manuring raises soil pH (Xu W. et al., 2023; Wang et al., 2024). In this experiment, crop type did not significantly affect pH, but biomass B amendments slightly decreased pH compared to the control. This acidifying effect was linked to altered Mg:K ratios: unamended soil had Mg:K ≈ 2.4, while biomass B-amended soils ranged from 1.4 to 1.8. Such shifts in cation ratios can influence soil buffering capacity and mineral availability (Jalali, 2008).

Although the magnitude of pH change was small, it appeared that even minor acidification may accumulate under repeated applications of green manures, especially when combined with shifts in Mg:K balance. This could alter root nutrient uptake and rhizosphere buffering, implying that crop maturity selection could serve as a management tool to fine-tune soil reaction for target crops.

Stoichiometric indicators revealed that radish at later harvest (RadB) intensified microbial C limitation (higher C acquisition and vector length) but reduced N acquisition, accompanied by greater N limitation (higher vector angle). In contrast, white mustard at later harvest (WmuB) promoted microbial N acquisition and alleviated N limitation. These findings suggest that the maturity stage of incorporated biomass is a decisive factor also in regulating soil nutrient availability and microbial resource allocation. Early incorporation enhances labile nutrient cycling, while later incorporation, particularly of white mustard, supports more balanced nutrient acquisition and may favour longer-term soil fertility.

In general, green manures are linked to higher soil C (Patra et al., 2023), however, in this study, no significant changes in oxidizable C were detected. Radish (lower plant C) degraded faster, whereas safflower (higher plant C) decomposed slowly, reducing its contribution to soil C enrichment. This illustrates a trade-off: C-rich biomass may not enrich soil if its decomposition is too slow to release labile substrates (Tamura et al., 2017; Gleixner et al., 2001). In general, the length of the experiment was too short to consider markable changes in total C contents that would be likely observed in similar medium- to long-termed trial.

Enzyme responses supported this view. β-glucosidase (GLU) activity was unchanged under biomass B but increased under RadA and SafA, coinciding with higher cellulose + hemicellulose contents and carbohydrate levels. Higher hemicellulose:ADL ratios in early harvests (A) compared to later harvests (B) likely explained these effects (Liu et al., 2025). Thus, enzyme activity was driven more by substrate quality than crop type alone.

C acquisition (Cacq.) and microbial C limitation (Vector length) responded only to biomass B, and most strongly in RadB. This biomass had high HCEL:ADL and Ash but low plant C and C:N, promoting greater microbial demand and C turnover. In contrast, soil P, K, and Mg generally increased with biomass amendments, consistent with other reports for P (Zhang et al., 2023; Kataria et al., 2024; Khan et al., 2025), potassium and magnesium (Solangi et al., 2019; Kucerik et al., 2024; Ma et al., 2021).

Interestingly, higher phosphatase activity (Phos) occurred in RadA and SafB, both associated with lower soil P content. This suggests microbial upregulation of P-mining under limitation, corroborated by higher vector angles (>45°) aligning with the notion about mutual relation between reduced availability of P and induced Phos (functioning also vice-versa (Bargaz et al., 2012; Janes-Bassett et al., 2022)).

Altogether, these results show that the maturity stage determines whether soil C inputs stimulate microbial activity (labile C cycling, possible priming of SOM) or contribute to persistent organic matter pools (slow decomposition, reduced turnover). The consistent link between stoichiometric indices and enzyme assays highlights that integrating these approaches is essential to reveal “hidden” nutrient limitations and microbial allocation trade-offs.

Correlation analysis confirmed strong links between plant biomass traits and soil enzyme dynamics. High C content in plant material was associated with increased urease activity and microbial N acquisition, while high Ash content showed the opposite trend, indicating that mineral-rich biomass may partially suppress N cycling. Fibre quality traits also played a key role: hemicellulose and lignin were strongly associated with residual biomass, suggesting that structural compounds drive decomposition resistance, while lignin negatively correlated with β-glucosidase, reflecting reduced microbial C acquisition from recalcitrant residues. Among soil properties, higher P, K, and Mg contents were linked with lower phosphatase activity, implying nutrient feedback regulation.

Contrary to some studies (Lyu et al., 2023; Wang et al., 2025), biomass incorporation did not significantly change soil TN or C:N ratios in this study. However, enzyme activities indicated shifts in N cycling. Urease activity increased under SafA, while WmuA and RadB showed reduced values compared to other treatments. This suggests that safflower may sustain N mineralization despite its slower decomposition.

N-acetyl-β-D-glucosaminidase (NAG), linked to fungal chitin turnover, increased in RadA and WmuB relative to the control. Crucifer biomass maturity negatively affected NAG, as seen in lower activity under SafB versus SafA, aligning with reports that later maturity reduces protein degradation (Kohn and Allen, 1995; Kejun et al., 2011). Arylsulfatase (ARS) also increased with biomass amendments, consistent with earlier findings (Tejada et al., 2008; Elfstrand et al., 2007). However, WmuA showed the lowest ARS, despite white mustard being rich in S-containing glucosinolates (Couëdel et al., 2018b). This discrepancy likely reflects that glucosinolate degradation is mediated by enzymes like myrosinase (Al-Turki and Dick, 2003; Huber et al., 1983) rather than ARS.

Taken together, these enzyme responses indicate that microbial N and S cycling can shift without parallel changes in bulk TN, highlighting the importance of process-level indicators. While safflower decomposes slowly, it still supports steady N turnover through urease activity. Crucifer residues, in contrast, showed maturity-dependent differences in protein- and chitin-degrading pathways, with direct implications for the timing of incorporation to optimize N release. Moreover, the interaction between glucosinolate-rich residues and sulfur enzymes suggests that biochemical specificity of plant metabolites must be considered when predicting S dynamics under green manuring.