Integrating winter legume cover crops in temperate rice cropping systems for enhancing carbon and nitrogen inputs without increasing GHG emissions

Transitioning to low-input cropping systems while enhancing soil fertility and limiting greenhouse gas (GHG) emissions is a major challenge for temperate rice cultivation. In northern Italy, cultivating a winter leguminous cover crop (CC) in conventional monocropped rice systems may represent an important source of organic matter (OM) to sustain soil fertility and nitrogen (N) for crop nutrition. However, incorporating fresh green manure before flooding may also stimulate methane (CH 4) production by increasing the availability of labile OM under anaerobic conditions. This study evaluated the agronomic and environmental implications of introducing Vicia villosa Roth as a winter CC in a rice-based system managed under dry seeding followed by continuous flooding. Plots with CC were compared to a winter fallow over two consecutive growing seasons, during which biomass inputs, porewater chemical composition, GHG fluxes, and soil organic matter (SOM) composition were evaluated. Hairy vetch established successfully in both years and provided inputs of 1.91 ± 0.32 and 2.72 ± 0.19 t C ha -1 and 180 ± 30 and 240 ± 30 kg N ha -1. Cumulative methane emissions over the entire rice growing cycle, as well as the global warming potential, showed no statistically significant differences between the treatments. Overall, these results suggest that integrating hairy vetch as a winter cover crop preceding dry-seeded temperate rice can substantially increase organic matter and N inputs without an evident trade-off in seasonal CH 4 emissions, thereby supporting the development of low-input and more environmentally sustainable rice paddy management strategies.

Temperate rice cropping systems face several challenges related to variable grain yields, low soil fertility and fertiliser use efficiency, and important environmental impacts, in particular greenhouse gas (GHG) emissions (Bhatia et al., 2023; Qian et al., 2023). In northern Italy, the main European district for rice cultivation, rice is typically grown in monoculture in the plains of the River Po where water availability allows for rice cultivation in flooded paddies (Miniotti et al., 2013). In this region, livestock-derived organic matter such as manure is generally scarce and soils often have relatively low soil organic matter (SOM) contents, making alternative strategies to increase organic inputs and improve fertility particularly relevant (Moretti et al., 2025). At the same time, rice cropping systems are also required to reduce their dependence on mineral fertilisers while sustaining grain yields. Rice production depends heavily on mineral nitrogen (N) fertilisers, but their mismanagement results in a typically low N use efficiency (NUE), higher costs, and serious environmental impacts (Farooq et al., 2022; Peng et al., 2010). Fertiliser N recovery in rice systems is often below 50% due to N losses via ammonia volatilisation, nitrate leaching, runoff, and denitrification (Farooq et al., 2022).

Introducing leguminous cover crops (CC) during the winter fallow period has been proposed as a promising strategy to increase organic carbon (C) and N inputs to the soil, favour SOM accumulation, improve structure and fertility, and reduce mineral N fertilisation (Lizcano Toledo et al., 2022; Sugai et al., 2024). In addition, cover crops provide off-season soil cover that can reduce nutrient losses and suppress weeds (Moretti et al., 2025). Hairy vetch (Vicia villosa Roth) is a leguminous cover crop well adapted to growing in temperate rice paddies, providing up to 100–130 kg N ha -1 y -1 and thereby contributing to rice N supply and grain yield while reducing mineral N inputs (Fogliatto et al., 2021; Lizcano Toledo et al., 2022). Delaying hairy vetch termination towards flowering generally increases biomass and nutrient inputs (Anugroho et al., 2009). However, incorporating fresh residues in proximity to paddy flooding also increases labile C and N in the soil, potentially promoting methane (CH 4) production under anoxic conditions and enhancing N loss (Bertora et al., 2018a). Residue incorporation timing is critical for CH 4 mitigation in paddy soils because it controls the degree to which fresh organic residues decompose aerobically before soil reduction occurs (Liu et al., 2025; Sander et al., 2014). Incorporation close to flooding enhances CH 4 production by supplying labile carbon directly under anaerobic conditions, promoting methanogenesis, while earlier incorporation under dry aerobic conditions, 10-30 days before flooding, reduces the amount of readily available C and therefore limits CH 4 emissions (Lee et al., 2020). Thus, identifying combinations of cover cropping and irrigation regimes that enhance SOM and nutrient retention without increasing CH 4 and nitrous oxide (N 2 O) emissions is essential for developing climate-smart rice systems (Hwang et al., 2017; Li et al., 2021). Previous studies on paddy systems have reported contrasting GHG responses, showing that the effect of green manuring on CH 4 emissions depends on cover crop species, biomass quantity and quality, residue management and the timing of flooding (Grohs et al., 2024; Hwang et al., 2017; Li et al., 2021).

Based on these considerations, we hypothesised that associating a winter leguminous CC with dry-seeded rice may provide important benefits in terms of organic matter and N inputs without significantly increasing CH 4 emissions, by allowing sufficient time for the aerobic decomposition of labile C inputs before field flooding at the tillering stage. The objective of this study was therefore to evaluate, under a commercial farming scenario of a typically monocropped temperate rice system, the benefits and trade-offs of introducing hairy vetch (Vicia villosa Roth) as a winter cover crop in combination with dry seeding of rice followed by continuous flooding. Specifically, we quantified (i) the contribution of leguminous green manure to OM and N inputs at the time of CC termination; (ii) the influence on SOM in terms of quality and quantity determined by the practice; and (iii) seasonal CH 4 and N 2 O fluxes and derived CH 4 emission factors, and growing-season global warming potential (GWP) for rice with and without a hairy vetch cover crop.

By combining agronomic, soil quality, and greenhouse gas assessments under real farm conditions, this study helps identify management practices capable of reconciling soil fertility improvement with climate mitigation in temperate rice systems. From this perspective, the findings may offer useful guidance for farmers, advisors, and policymakers.

Study site

The study was carried out between autumn 2022 and autumn 2024, covering two rice-growing seasons and their preceding winter fallow period, in a commercially managed monocropped rice field situated in the plain of the River Po (north-western Italy; 45°12′10.7″ N, 8°40′27.5″ E) within one of Italy’s main rice-growing districts. The area has a temperate sub-continental climate, with peak rainfall typically occurring in spring and autumn. Mean daily air temperature and cumulative precipitation recorded during the experimental period are reported in Figure 1. The soil is an Inceptic Haplustalf, coarse loamy, mixed, super active, mesic (Soil Survey Staff, 2022) with a topsoil (Ap horizon, 0–20 cm) having a sandy loam texture, a pH in water of 6.1 (soil: solution ratio 1:2.5), SOC content of 14.1 g kg -1, total nitrogen (TN) of 1.5 g kg -1, and a C/N ratio of 9.1.

Experimental design and field management

The hydrologically isolated commercial paddy field under investigation was split into two sub-plots, each having an area of 5,541 and 6,679 m², respectively, and hosting two contrasting winter management practices. Before the establishment of the trial in 2022, the field had been managed as a single homogeneous unit, without subdivision into sub-plots, and had been continuously cultivated with rice under the same agronomic management, with no previous use of winter cover crops. In one sub-plot, a leguminous cover crop for green manure was grown during the winter, and the biomass was incorporated into the soil in spring together with rice straw residues remaining from the previous crop (CC). The second sub-plot was kept fallow during the winter, and rice straw was left on the soil surface after harvest until spring tillage (No CC). Apart from tillage operations, water management and rice cropping were identical in both sub-plots and consisted of dry-seeding followed by continuous field flooding from the tillering stage, approximately one month after sowing, until early maturation in August when the field was drained to allow for grain ripening and harvest operations (Supplementary Table 1).

The cover crop Vicia villosa Roth (hairy vetch) was selected for its high biomass production (Parr et al., 2011). In both years, vetch was sown in autumn after the rice harvest (10 October 2022 and 15 October 2023). Cover crop termination at the flowering stage and biomass incorporation were carried out by ploughing. The rice cultivar selected by the farmer was Oryza sativa var. Omega CL ® (round grain variety), drill-seeded on May 16, 2023 and May 20, 2024, approximately 2 to 3 weeks after cover crop termination. At the tillering stage, both sub-plots received fertilisation comprising 70 kg N ha -1 in 2023 and 50 kg N ha -1 in 2024, together with 50 kg K ha -1 without any phosphorus fertilisation, and were subsequently flooded, maintaining a ponding water depth of 10–15 cm. A second top-dressed N fertilisation dose was applied to flooded soils at panicle initiation, comprising 40 kg N ha -1 in 2023 and 30 kg N ha -1 in 2024, for a total supply of 110 and 80 kg N ha -1 in the first and second years, respectively. The field was permanently drained in late August to allow for rice maturation. Rice harvest took place in October in both seasons. Additional agricultural management details are reported in Supplementary Table 1.

Vetch and rice production

Immediately before CC termination, hairy vetch aboveground biomass was determined by cutting vegetation at ground level from three spatially representative 0.25 m² quadrats per sub-plot. Rice straw and grain yields were sampled similarly at harvest. All biomass samples were oven-dried at 60 °C to a constant weight to determine dry matter and ground to a fine, homogeneous powder before elemental analysis. Total C and N contents were determined by elemental analysis (UNICUBE, Elementar, Germany). Organic C and total N inputs from green manuring were calculated as the product of aboveground dry matter and their element contents. Nitrogen uptake by the rice was estimated separately for grain and straw as the product of harvested biomass and corresponding N content.

The short-term apparent contribution of hairy vetch N to rice N uptake was evaluated annually using two descriptive indices: grain organic N use efficiency (ONUEg) and total aboveground organic N use efficiency (ONUEt), calculated as follows:

Where N grain CC and N grain No CC are grain N uptake in the CC and No CC treatments, respectively; N straw CC and N straw No CC are straw N uptake in the two treatments; and N CC biomass is the N input supplied by hairy vetch aboveground biomass at termination. The indices were calculated on an annual basis using treatment means and are therefore intended as descriptive indicators of apparent N recovery rather than direct estimates of cover crop-derived N uptake.

Soil porewater analyses and gas flux measurements

Two monitoring stations were established in each treatment sub-plot. To minimise edge effects, both stations were located in the central portion of the sub-plot, approximately 10 m from the levees. The two stations within each sub-plot were separated by approximately 90 m, while the minimum distance between stations in different treatment subplots was 70 m. Each monitoring station included two gas-flux chambers and two porewater samplers. Monitoring stations were used to characterise within-plot spatial variability (n = 4; Figure 2). Porewater samplers (10 cm length, 0.15 µm pore size; MacroRhizon, Rhizosphere Research, Wageningen, The Netherlands) were installed vertically at a depth of 15 cm to collect soil solution above the plough pan, with four sampling locations per plot. Porewater was collected weekly from rice seeding until final field drainage. Approximately 20 mL of pore water was extracted per sampling time and analysed for dissolved Fe(II) and dissolved organic carbon (DOC). Dissolved Fe(II) concentrations were quantified colorimetrically immediately after sampling, using the 1,10-phenanthroline method (Loeppert and Inskeep, 1996). DOC concentrations were determined by Pt-catalysed high-temperature combustion (850 °C) followed by infrared detection of CO 2 (VarioTOC, Elementar, Hanau, Germany), after removal of inorganic carbon by acidification to pH 2.

Gaseous CH 4 and N 2 O fluxes were measured weekly from rice seeding until final field drainage, at four chamber sampling locations per treatment plot, using the non-steady-state closed chamber technique (Bertora et al., 2018b). A few days after sowing the rice, stainless steel anchors (75 cm × 36 cm × 40 cm) were inserted into the soil to a depth of 40 cm. Wooden planks were installed to access the measurement stations from the paddy levees without disturbing the surrounding soil. During each measurement, a rectangular stainless-steel chamber (75 cm × 36 cm × 20 cm high) was sealed to each anchor. To limit temperature variations during measurement, chambers were covered with 5-cm thick light-reflective insulation and were equipped with a pressure vent valve as described by Bertora et al., 2018b. Gas samples were collected at four time points after chamber closure (0, 10, 20, and 30 min) and transferred to 12 mL evacuated vials closed with butyl rubber septa (Exetainer ® vial from Labco Limited, UK). Gas concentrations in collected samples were determined by gas chromatography using a fully automated gas chromatograph (Agilent 7890A with a Gerstel Maestro MPS2 autosampler) equipped with electron capture, thermal conductivity, and flame ionisation detectors for the quantification of N 2 O and CH 4, respectively.

CH 4 and N 2 O fluxes were estimated from chamber concentration time series using the R package gasfluxes (Fuss, 2016), by fitting multiple models (linear, robust linear, Hutchinson–Mosier regression (HMR), and non-steady-state diffusion estimator (NDFE)). The final flux estimate was selected using the kappa.max criterion implemented in selectfluxes(), based on the HMR curvature parameter κ (Hutchinson and Mosier, 1981). Cumulative growing-season emissions were calculated by linearly interpolating fluxes between sampling dates.

To compare the climatic impact of greenhouse gas (GHG) emissions and assess system efficiency, the global warming potential (GWP) and eco-efficiency were calculated. The GWP was estimated by applying IPCC 100-year characterisation factors of 27 and 273 to convert cumulative CH 4 and N 2 O emissions to CO 2 equivalents, respectively (IPCC, 2023). Eco-efficiency (kg grain yield kg -1 CO 2 eq) was defined as the ratio between rice grain yield and GWP, representing the amount of grain produced per unit of GHG emissions. Emission factors (EFs) for CH 4 emissions associated with each treatment (EF CH 4 expressed as kg CH 4 ha -1 d -1) were determined by dividing total cumulative CH 4 emissions over the rice cropping period, from seeding to harvest, by the duration of the crop cycle (143 days in 2023 and 163 days in 2024).

Soil sampling and analyses

Topsoil samples were collected at the end of the second year using a random spatial scheme within each sub-plot. Three composite samples were collected, each obtained by pooling three subsamples to describe spatial variability within each treatment plot. Soil organic carbon (SOC) and total nitrogen (TN) were determined by dry combustion using an elemental analyser (UNICUBE, Elementar, Germany). At the same time, density fractionation was employed to separate and quantify three SOM functional pools: free particulate organic matter (fPOM), aggregate-occluded particulate organic matter (oPOM), and mineral-associated organic matter (MAOM), using sodium polytungstate (1.6 g cm -3) and sequential centrifugation (12,800 g for 20 min) before (fPOM) and after (oPOM) ultrasonication (175 J mL -1, <40 °C). The residual heavy material was recovered as MAOM after repeated washing and centrifugation (≥ 17,700 g, ≥ 30 min) (Cerli et al., 2012; Colombatto et al., 2026). All fractions were dried (40 °C), homogenised, and analysed for organic C and total N by dry combustion.

Statistical analysis

Biomass variables (hairy vetch aboveground biomass and associated C and N inputs, rice grain and straw yield, C and N contents, and N uptake) were reported as means ± standard deviation (SD) of within-plot subsamples. Subsamples, collected within each plot, were treated as descriptive subsamples rather than as independent treatment replicates. Additionally, exploratory Welch’s t-tests were calculated for rice grain and straw yield and N uptake to compare subsamples between treatment plots within each year; these p-values are reported only as an aid to interpretation. Differences in hairy vetch biomass and associated C and N inputs between years were interpreted descriptively. For variables measured repeatedly over time, such as CH 4 and N 2 O fluxes, dissolved Fe(II) and DOC, treatment effects were evaluated using linear mixed-effects models (lme4 package; Bates et al., 2015), with treatment, sampling date, and their interaction as fixed effects, and replicate as a random intercept. Residual diagnostics were inspected for all models. For porewater Fe(II) and DOC, these diagnostics were used to choose the scale of analysis: Fe(II) and DOC in 2023 were transformed using the natural logarithm scale, whereas 2024 data were analysed on the original scale. Daily CH 4 and N 2 O flux distributions were strongly right-skewed with many small values and occasional peaks; for this reason, both CH 4 and N 2 O fluxes were analysed on the natural logarithm scale in both years. Even after transformation, residual diagnostics indicated deviations from perfect normality and homoscedasticity; therefore, these models were interpreted as approximate tools to describe temporal patterns and treatment differences within the field studied. Seasonal cumulative CH 4 and N 2 O emissions were analysed using linear mixed-effects models with year, treatment, and their interaction as fixed effects and replicate as a random effect. GWP was analysed using linear models with year, treatment, and their interaction as fixed effects. Eco-efficiency, because yield and GWP were measured on independent experimental units with different replication levels, was reported descriptively, with uncertainty quantified via the delta method, and no inferential tests were applied to it. Organic C and total N and SOM fractions (fPOM, oPOM, MAOM) were compared between treatments using Welch’s t-tests. The results were considered to be statistically significant when p < 0.05.

Vetch and rice production

The establishment of hairy vetch was successful in both years with mean plant densities, assessed between late November and early December, of 46 and 84 plants m -2 in 2022 and 2023, respectively. Greater aboveground biomass at flowering was produced in 2024 than in 2023, resulting in higher organic C and total N inputs with green manuring in 2024 (Table 1). Incorporated CC biomass quality was similar between years, with C/N ratios approximately 11 in both years.

Despite the additional N inputs from hairy vetch and although field observation suggested greater visual vigour in the CC treatment during the reproductive stages, rice grain yields were similar between CC and No CC plots in both years, indicating no clear treatment differences within each year (p > 0.05; Table 2). Grain N uptake was slightly higher but still not significant in the CC plot in both years (p > 0.05; Table 2). Rice straw showed a similar pattern: biomass and C content were comparable between treatments, whereas straw N concentration and N uptake also tended to be higher in the CC plot, again without clear between-treatment differences (p > 0.05; Table 2).

In the CC treatment, the contribution of cover crop N to rice N uptake, quantified using ONUE indices, showed high variability between the 2 years. The ONUEg indicator varied between 10% and 4% in 2023 and 2024, respectively, while ONUEt ranged from 13% to 19%.

Porewater chemistry and GHG emissions

Changes in porewater chemistry and GHG emissions over time were driven mainly by water management rather than by CC incorporation in both years. Porewater Fe(II) concentrations were generally low during the initial aerobic phase and increased markedly after the onset of field flooding, reaching maximum values of 67 mg Fe L -1 around mid-season before declining towards final drainage (Figure 3a). Fe(II) concentrations were overall higher in 2024 than in 2023 and increased progressively until final drainage, whereas in 2023 they remained lower and showed a more irregular seasonal pattern. Concentrations were similar between CC and No CC treatments in both years, and the mixed-effects model detected no statistical differences between treatments.

Porewater DOC concentrations were also strongly influenced by water management, increasing after the onset of flooding and declining again towards final drainage (Figure 3b). Porewater DOC concentrations showed similar trends to those observed for dissolved Fe(II) and were overall higher in 2024 than in 2023. In 2024, DOC followed a more regular seasonal pattern, increasing progressively throughout the flooding period until final drainage. DOC dynamics were largely comparable between treatments in both years (p > 0.05), and no consistent CC effect on DOC concentrations was noted across years.

In both years, daily CH 4 fluxes followed a typical pattern for flooded rice, with negligible emissions before flooding, a sharp increase after field flooding, maximum peaks between stem elongation and booting approximately 50 and 70 days after seeding, and a gradual decline towards maturity and after final drainage (Figure 4a). Treatment effects on CH 4 fluxes were very limited: the mixed-effects model identified significant differences between CC and No CC only on isolated sampling dates in 2024 (p < 0.05; Figure 4a), where treatment without CC tended to exhibit slightly higher fluxes. N 2 O fluxes were characterised by short-lived peaks in proximity to mineral fertiliser applications at the beginning of the cropping season, when the field was not yet flooded, and were subsequently undetectable, close to zero, during most of the flooded period (Figure 4b). Significant treatment differences occurred only for early-season peaks (p < 0.05), when the CC treatment showed higher N 2 O emissions than the control; however, no sustained treatment effect was observed during the flooded phase when fluxes were low and sometimes below the limits of detection.

Seasonal cumulative CH 4 emissions were not significantly affected by treatment, year, or their interaction (p > 0.05). Mean cumulative CH 4 emissions were 229.3 ± 80.9 kg CH 4 ha -1 under CC and 288.2 ± 126.3 kg CH 4 ha -1 under No CC (Figure 5a). Similarly, the derived EF CH 4 were not significantly affected by treatment, year, or their interaction (p > 0.05) and ranged from 1.50 ± 0.54 to 1.87 ± 0.75 kg CH 4 ha -1 d -1 for CC and No CC, respectively. Cumulative N 2 O emissions did not differ significantly between treatments in either year (p > 0.05). However, in both years the CC treatment tended to show slightly higher N 2 O losses, although differences remained non-significant, with cumulative emissions of 8.8 ± 11.4 and 2.2 ± 0.6 kg N 2 O ha -1 in 2023 and 2024, respectively, compared to 6.8 ± 7.7 and 0.3 ± 0.3 kg N 2 O ha -1 in No CC.

Consequently, the total GWP of the two treatments, expressed as kg CO 2 equivalents ha -1, did not differ significantly between treatments in either year, being on average 8200 kg CO 2 equiv. ha -1 (p > 0.05; Figure 5c). Methane was the main contributor to the GWP, accounting on average for about 86% of the total. Eco-efficiency values were around 0.74 kg grain kg -1 CO 2 equiv. with slightly lower values for No CC with respect to CC (Figure 5).

Soil organic C and functional pools

After 2 years of different winter management, SOC and TN in the 0 – 20 cm layer did not show any significant differences between treatments (mean values 13.17 ± 1.82 g C kg -1 in CC and 12.30 ± 1.56 g C kg -1 in No CC; and 1.53 ± 0.21 g N kg -1 in CC and 1.53 ± 0.21 g N kg -1 in No CC, p > 0.05). However, the soil C/N ratio was significantly lower in the No CC compared to the CC treatment (8.03 and 8.59, respectively, p = 0.012). Across treatments, most SOC and TN were recovered in the MAOM fraction (88.8% and 93.4%, respectively), while fPOM and oPOM accounted for smaller, yet non-negligible pools, representing on average 2.7% SOC, 1.7% TN, and 8.5% SOC, 4.9% TN, respectively (Figures 6b, c). No significant differences were observed in the distribution of C and N in the three fractions between treatments. Similar to bulk soil results, the C/N ratio of the MAOM fraction was significantly lower in No CC than in the CC treatment (p < 0.05).

Cover crop effects on C and N inputs

The incorporation of hairy vetch through green manuring led to a substantial additional input of N and organic matter into the system, although rice grain and straw production remained broadly similar between treatment plots across years, even though the crop appeared more vigorous in the presence of CC during the reproductive stages. Hairy vetch established successfully in both years, although biomass production showed some interannual variability. In fact, the earlier establishment and favourable spring rainfall in 2024 supported higher biomass production, with hairy vetch reaching 6.14 ± 0.47 t ha -1 dw at flowering (Table 1). Termination at flowering, which generally corresponds to peak plant N content (Anugroho et al., 2009), resulted in N inputs of between 175 and 242 kg N ha -1. These values are consistent with previous results reported by Lizcano Toledo et al. (2022) in the same area and are comparable to or even higher than conventional mineral N fertilisation doses in flooded rice systems in northern Italy (about 120–130 kg N ha -1; Zavattaro et al., 2008), confirming the potential of legume cover crops to partially replace fertiliser N while increasing OM inputs to the soil (Thapa et al., 2022; Zhang et al., 2025). The incorporation of hairy vetch contributed a modest amount of OC (1.9-2.7 t C ha -1), in line with previous studies reporting a mean supply of 1.4-2.1 t C ha -1, thereby representing an additional source of SOM that may support soil structural stability and long-term SOM maintenance (Haque et al., 2020; Lizcano Toledo et al., 2022; Sugai et al., 2024).

Despite these inputs, rice grain yields remained broadly similar between CC and No CC plots across years, indicating that the additional N supplied by hairy vetch did not translate into a clear short-term yield advantage under the experimental conditions of this study (Table 2). A possible explanation is that mineral fertilisation already provided sufficient N to meet crop demand such that CC-derived N represented an additional input in excess of yield requirements. Excessive N supply in dry-seeded rice does not necessarily increase yield and may instead favour vegetative growth, unproductive tillering, and reduced dry matter translocation to the panicle (Zhen et al., 2026). This interpretation is consistent with previous evidence showing that the yield benefits of cover crops in dry-seeded rice are more evident when mineral N application rates are reduced (≤ 80 kg N ha -1; Lizcano Toledo et al., 2022). In 2024, the greater visual vigour observed in the CC plot during the reproductive stages did not translate into higher grain yield, suggesting that additional vegetative growth was not fully converted into grain production, possibly due to constraints on dry-matter translocation or grain filling during the final maturation phase, which may have been exacerbated by the cooler temperatures recorded in September 2024. In addition, long-term studies on green manuring in rice systems indicate that improvements in crop productivity are often associated with the cumulative effects of repeated green manure incorporation and partial substitution of chemical fertilisers, rather than with immediate yield responses after short-term adoption (Zhang et al., 2023). However, yield responses should still be interpreted cautiously, as field-scale grain production is influenced by several uncontrolled agronomic and environmental factors.

In the CC treatment, the organic N use efficiency indices (ONUEg and ONUEt) indicated that only a modest fraction of vetch-derived N was recovered in grain in the short term, whereas a larger share was recovered in total aboveground biomass. The remaining CC-derived N was likely distributed between belowground residues, SOM fractions and loss pathways (Sugai et al., 2024; Yang et al., 2025). The relatively low ONUE values and the absence of a clear yield response to CC should therefore be interpreted in relation to both N availability and dry-seeded crop establishment. In this study, mineral N fertilisation was reduced in 2024 compared to 2023 (110 and 80 kg N ha -1, respectively), possibly explaining the slightly higher apparent short-term recovery of cover crop-derived N in the total aboveground biomass, albeit without a significant yield benefit. Moreover, under aerobic dry seeding, part of the N released during the decomposition of hairy vetch residues may have become available before the crop was able to use it, particularly if rice establishment was constrained by limited soil water availability. Indeed, cover crops can reduce soil water availability at sowing of the subsequent crops, potentially affecting crop establishment and yield response (Garba et al., 2022).

These mechanisms may help explain why the additional N supplied by hairy vetch did not translate into higher grain yield in the short term. Overall, our results suggest that, in systems combining winter legume cover crops and dry-seeded rice, mineral N management may need to be adjusted, for example by reducing mineral N rates or modifying fertiliser timing, to improve synchrony between N availability and crop demand. Future studies should also evaluate whether a mixture of legume and grass cover crops could slow early N mineralisation by increasing residue C/N ratio and therefore improve the synchronisation between cover crop N release and rice N uptake.

Short-term response of SOM fractions to CC

The effects of green manuring on bulk topsoil SOC and TN after only two rice-growing seasons were limited and in line with the short duration of the study, despite the substantial additional C and N inputs supplied with the incorporation of the hairy vetch biomass. This agrees with long-term studies showing that detectable increases in SOC under cover crops usually emerge only after several years of continuous management, especially in relatively coarse-textured soils and with high-quality, low C/N residues that favour rapid mineralisation rather than short-term C accumulation (Hu et al., 2023; Thapa et al., 2022).

The separation of SOM into functional pools by density fractionation provided further insight into SOM quality. In both treatments, the mineral-associated organic matter (MAOM) fraction constituted the largest proportion of soil C and N, while free and aggregate-occluded particulate fractions (fPOM, oPOM) represented smaller pools (Mayer et al., 2022; Zhang et al., 2022). Contents of C and N in the individual fractions did not differ significantly between treatments, but the C/N ratio of MAOM was slightly but significantly lower in the No CC treatment, whereas particulate fractions showed no marked differences (Figure 6c). This result is somewhat counterintuitive, as the incorporation of N-rich hairy vetch residues could be expected to decrease, rather than increase, the C/N ratio of the mineral-associated pool. However, this pattern suggests that hairy vetch may have already altered the qualitative composition of the mineral-associated pool rather than its size (Jilling et al., 2020). On the one hand, inputs of high-quality leguminous residues with a low C/N ratio may promote microbial C use efficiency, leading to the formation of microbially processed organic matter and stabilisation of necromass on minerals, potentially increasing C in MAOM without a proportionate increase in N in MAOM (Cotrufo et al., 2013; Connell et al., 2025). On the other hand, the fraction of N in more labile forms can be rapidly removed through plant uptake, gaseous losses and leaching, leading to slower net accumulation of N relative to C in MAOM under alternating aerobic and anaerobic conditions. Recent work also suggests that MAOM can contribute substantially to potentially mineralisable N and plant-available N, challenging the view of this fraction as strictly inert (Villarino et al., 2023). Longer-term monitoring will be necessary to determine whether these shifts translate into measurable increases in MAOM C and total SOC as observed in multi-year experiments with cover crops and organic farming systems (Thapa et al., 2022).

Cover crop incorporation did not enhance GHG emissions under dry-seeded rice

Seasonal dynamics of porewater chemistry and gas fluxes indicate that in the context of the dry-seeded rice cropping systems investigated here, water management rather than cover crop incorporation itself was the main driver of biogeochemical conditions controlling the establishment of reducing conditions and the availability of labile C during the rice season. In both treatments, porewater Fe(II) concentrations remained low during the initial aerobic phase, increased after flooding, and then declined towards final drainage, as commonly observed in flooded paddy soils (Said-Pullicino et al., 2016). The absence of systematic treatment differences in Fe(II) suggests that, despite the larger labile organic C inputs, hairy vetch cropping and residue incorporation did not substantially alter the onset or intensity of reductive processes in the soil. Likewise, DOC concentrations were generally similar between CC and No CC (Figure 3b), indicating that cover crop incorporation did not increase the pool of soluble, readily available C at the onset of field flooding. Fresh residue incorporation and its timing with respect to paddy field flooding have been often associated with the rate of establishment of reduced conditions, the dissimilatory reduction and dissolution of Fe(III) and Mn(IV) oxyhydroxides in the absence of O 2, the desorption of soil-derived organic C previously stabilised by these minerals, the accumulation of porewater DOC, and the increase in substrate availability for methanogenesis, leading to higher CH 4 emissions (Bertora et al., 2018a; Ye and Horwath, 2017; Said-Pullicino et al., 2016). In our case, however, hairy vetch residues were incorporated several weeks before flooding to allow for dry seeding, so most of the easily degradable C was likely mineralised under aerobic conditions, favouring rapid CO 2 release rather than the accumulation of soluble DOC in porewater and subsequent CH 4 production (Lee et al., 2020). In addition, long-term adoption of leguminous green manuring has been shown to favour the stabilisation of C in more persistent, mineral-associated forms and to increase the aromaticity and molecular weight of dissolved organic matter, rather than preserving it as highly labile DOC (Gao et al., 2018; Huang et al., 2025).

Consistent with the interpretation of porewater chemistry, daily CH 4 fluxes followed the expected seasonal pattern for flooded rice, with negligible emissions before flooding, a sharp increase after field flooding, a mid-season peak around booting stage, and a decline towards final drainage (Figure 4a; Kim et al., 2013; Qi et al., 2024). Green manuring did not induce a consistent increase in CH 4 fluxes, and significant treatment effects were detected only on a few individual sampling dates.

Seasonal cumulative CH 4 emissions exhibited interannual variability but no clear and consistent effect of cover crop incorporation (Figure 5a), in line with previous evidence that the time interval between residue incorporation and field flooding is a major determinant of methanogenic stimulation in rice paddies (Kim et al., 2015; Lee et al., 2020). Studies in which CC residues are incorporated immediately before or under continuous flooding often lead to marked increases in CH 4 emissions due to enhanced supply of labile OM for methanogens (Haque et al., 2013; Kimura et al., 2004). However, delayed flooding associated with dry seeding may have allowed most of the easily degradable C to decompose under aerobic conditions, thereby limiting CH 4 production during the rice season (Bertora et al., 2018a). It has been previously reported that 10 days of aerobic degradation of the incorporated cover crop biomass before flooding can result in a 60% reduction in CH 4 emissions compared to flooding immediately after incorporation, and extending this period further leads to an even greater mitigation potential (Lee et al., 2020). Consistent with this interpretation, our findings support the hypothesis that dry seeding may provide a sufficiently long time interval between hairy vetch incorporation and flooding, during which aerobic soil conditions persist, thereby partially separating fresh C inputs from periods of maximum methanogenic activity (Bertora et al., 2018a; Peyron et al., 2016). The agronomic implication of this finding suggests that the association between winter cover crops and dry-seeded rice in temperate cropping systems may allow for delaying cover crop termination in order to increase N inputs without increasing CH 4 emissions.

N 2 O fluxes were characterised by short-lived peaks associated with residue incorporation and fertiliser application during the pre-flooding aerobic phase, followed by near-zero emissions once the field was fully flooded (Figure 4b). This pattern reflects the combined effects of cover crop incorporation and irrigation regime on N transformations. Hairy vetch likely increased mineral N availability and supplied labile organic matter, thereby creating conditions that could support both nitrification and denitrification during the aerobic or weakly reducing phases (Hu et al., 2021). However, once flooding is fully established, sustained anaerobic conditions suppress nitrification, promote nitrate depletion, and favour complete denitrification to N 2, thereby limiting N 2 O emissions (Peyron et al., 2016). CC treatment occasionally showed slightly higher early-season peaks, consistent with greater N availability and the potential for enhanced nitrification of mineralised N from vetch residues. It is important to note that N 2 O emissions are highly intermittent and strongly influenced by small-scale management and soil heterogeneity, which makes both measurement and upscaling uncertain (Ju et al., 2024). Moreover, green manure adds spatial complexity by creating localised hot spots of N, so chamber placement and sampling frequency can have a disproportionate influence on estimated N 2 O emission magnitudes (Yao et al., 2024).

Nonetheless, cumulative N 2 O emissions remained small compared with CH 4 and did not differ significantly between treatments (Figures 5b, c). Consequently, growing-season global warming potential (GWP) was strongly dominated by CH 4, with N 2 O contributing only marginally, as commonly reported for other flooded rice systems (Vitali et al., 2024). Consistent with the absence of treatment effects on GWP, eco-efficiency (grain yield per unit of CO 2 equivalent emitted) was broadly similar between treatments within each year (Figure 5c). Overall, results indicate that introducing hairy vetch in dry-seeded rice cropping systems did not worsen the climate performance of the system relative to a winter fallow. However, a replicated factorial experiment comparing seeding method (dry seeding vs. water seeding) and winter fallow management (CC vs. No CC) would be needed to directly test whether dry seeding affects emissions.

Implications for CH 4 emission factors

Beyond their agronomic relevance, these results are important for GHG accounting because they highlight a mismatch between measured CH 4 emissions and the assumptions embedded in the default Tier 1 factors proposed by the Intergovernmental Panel on Climate Change (IPCC) for estimating CH 4 emissions. In fact, for large-scale GHG accounting, the IPCC Tier 1 approach is based on default CH 4 emission factors (EF CH 4), defined as the average daily CH 4 emission per unit area during the rice-growing period (kg CH 4 ha -1 d -1), applied to national rice cultivation areas (Buendia et al., 2019). This framework is easy to use but provides only broad estimates because it does not capture the effects of specific management practices or local soil and climatic conditions. By applying the Tier 1 framework, a temperate rice paddy managed with a continuously flooded irrigation regime during the cultivation period, preceded by a non-flooded off-season fallow period of >180 d would be attributed a water regime scaling factor of SF w = 1.00 and SF p = 0.89, respectively, yielding a default daily EF CH 4 of around 1.4 kg CH 4 ha -1 d -1 (Buendia et al., 2019). The cultivation of a winter cover crop and incorporation of a green manure would further increase the Tier 1 estimate via the organic-amendment scaling factor SF o. Based on the measured average cover crop OM input of 5.45 t ha -1 dw, applying the IPCC Tier 1 CFOA factor of 0.45 for green manuring would result in an SF o = 2.08, equivalent to a doubling of the total annual CH 4 emissions over the cropping season with respect to the winter fallow practice, regardless of the type of seeding adopted (Buendia et al., 2019).

In this study, mean EF CH 4 between 1.50 ± 0.54 and 1.87 ± 0.75 kg CH 4 ha -1 d -1 were obtained with no significant differences between CC treatments, suggesting that in systems where winter cover crops are combined with dry seeding and delayed flooding, default Tier 1 factors may overestimate the incremental CH 4 emissions attributed to green manuring. Measured cumulative CH 4 emissions are consistent with IPCC 2019 Tier 1 defaults for continuously flooded rice in Europe without organic amendments (baseline EF CH 4 = 1.56 kg CH 4 ha -1 d -1; Buendia et al., 2019), as well as with previous values reported by Bertora et al. (2018a) for dry seeded Italian rice paddies with rice crop residues as the only OM input (EF CH 4 = 1.57 kg CH 4 ha -1 d -1). The observation that introducing a hairy vetch cover crop before a dry-seeded rice crop in this scenario did not incur additional CH 4 emissions (i.e., SF o ≈ 1), highlights a potential limitation in the adoption of the IPCC Tier 1 methodology for this specific management combination: water regime scaling factors distinguish continuous vs. intermittently flooded systems but do not explicitly account for dry seeding and the possibility that incorporated cover crop residues may decompose rapidly before the field is flooded without contributing additional CH 4. Based on these considerations, although we cautiously support the adoption of an SF o = 1 for a more appropriate CH 4 emission accounting for this combined practice in climate-policy frameworks for temperate rice (IPCC, 2023), further empirical support and field data are warranted as other factors (e.g., seasonal temperature and precipitation patterns, soil properties, and agronomic factors such as cover crop termination practice) could play an important role in determining the trade-off between green manuring and GHG emissions.

This study shows that integrating a hairy vetch leguminous cover crop during winter into dry-seeded temperate rice cropping systems can substantially increase soil organic C and N inputs without a corresponding increase in seasonal GHG emissions. Cover crop performance varied strongly between years, and the short-term translation of additional N into higher grain yields remained limited, even though green manuring represents an N input of the same order of magnitude as conventional mineral N fertiliser doses. These findings are particularly relevant for mono-cropped rice systems of northern Italy, where access to livestock-derived organic amendments is often limited and where reducing dependence on mineral N fertilisers is a major agronomic and environmental goal. Future work should therefore address longer time series, broader environmental gradients and true field replication in order to better resolve the agronomic payback time of this practice and its implications for SOC accumulation, N cycling and GHG balances.