The experiment was conducted in a typical Karst area in Machang town (26°25′N, 106°27′E), Guiyang, China. The region possesses a subtropical humid monsoon climate. The previous crop was maize, which is fallow in winter. The maize growing season is from April to September, with precipitation of 932.60 and 784.00 mm and mean temperatures of 20.35 and 21.09°C in 2020 and 2021, respectively. The soil type is yellow loam. The basic soil pH was 5.28, and the organic matter, total N and P were 34.94, 1.30, and 1.24 g kg⁻¹, respectively.

In order to choose the suitable green manure to improve the maize yield, NUE, and PUE, different green manure treatments, including (1) winter fallow field + no fertilization (CK); (2) winter fallow field + traditional chemical fertilizer (WF); (3) alfalfa + traditional chemical fertilizer (AL); (4) common vetch + traditional chemical fertilizer (CV); and (5) oilseed rape + traditional chemical fertilizer (OR), were established in randomized plots (3 m × 6 m) with a plot spacing of 0.3 m. There were three replicates (plots) for each treatment. The traditional chemical fertilizer application rate was N 244 kg ha⁻¹ and P₂O₅ 145 kg ha⁻¹.

Alfalfa, common vetch and oilseed rape were sown after maize harvest in 2019 and 2020 by broadcast seeding, with seeding amounts of 27, 45, and 45 kg ha⁻¹, respectively. No fertilizers were applied during the green manure growing season. All green manures were cut into 5–10 cm pieces and incorporated into the tillage layer (20 cm) 1 week before planting maize. The properties of green manures are shown in Table 1.

The maize varieties were starnuo41 and Yingtai863A in 2020 and 2021, respectively. All the seeds were planted in hill-holes, with three–four seeds per hill-hole, with a distance between the holes of ~60 cm. After the emergence of seedlings, each hole contained two plants. Monoammonium phosphate fertilizer (including N 11%, P₂O₅ 44%) was applied as a base fertilizer on the day of sowing maize, and urea (N 46%) as the top fertilizer was applied at the seedling stage (June 15, 2020, June 2, 2021) and before plucking and silking (July 10, 2020, July 6, 2021), accounting for 35 and 55% of the total N amount, respectively. Other management practices were the same as local conventional field practices.

Three topsoil (0–20 cm) samples were collected simultaneously with the maize harvest period in each plot, mixed evenly using the quartic method, then divided into two subsamples. One subsample were sieved through 2 mm sieves and stored at −20°C, and another subsample was air dried and sieved through 0.15 and 0.25 mm sieves when relevant soil properties were measured. At the maize harvest stage, ten maize plants were randomly harvested in each plot to measure grain yield and stem and leaf biomasses. The stem and leaf samples were dried for 30 min at 105°C and then dried for 48 h at 60°C. The dried grain, stem, and leaf were sieved (0.15 mm) to determine their N and P contents.

NH4−-N and NO3+-N were extracted with 1 M KCl (1:5 w/v) and measured with a flow analyzer (Cleverchem 380, Germany). Soil AN content contained NH4−-N and NO3+-N contents. The total N content was measured by the Kjeldahl digestion procedure (Bremner and Tabatabai, 1972). The soil total P content was determined by the molybdenum blue method after digestion with concentrated HClO₄-H₂SO₄ (1:10 v/v), and the soil AP was determined via the molybdenum blue method after extraction with 0.5 M NaHCO₃ at pH 8.5.

The maize N or P uptakes of grain (Eq. 1), stem (Eq. 2), leaf (Eq. 3), and straw (Eq. 4) were calculated as follows:

NUE (Eq. 5) and PUE (Eq. 6) were determined as follows (Cao et al., 2021):

where F denotes the amount of N and P applied (kg ha⁻¹), U_+NorP and U_−NorP are the total N and P uptake by grain and straw in the chemical fertilizer treatment and CK (kg ha⁻¹), respectively.

The LSD test was used to test the differences between the yields, N and P concentrations in maize plants, N and P uptake, soil available N and soil available P content among treatments at P < 0.05, and a t-test was used to test the differences in NUE and PUE at P < 0.05. Two-way ANOVAs were used to determine the effects of green manure type, year, and their interaction on maize yield and nutrient use efficiency at P < 0.05. The “piecewise SEM” package in R (v4.1.1) was used for structural equation modeling (SEM). The a priori model included all possible pathways among these factors. To reduce the complexity of SEM, the representing indices of green manure N and P concentrations were calculated using PCA with the “FactoMineR” package in R (v4.1.1). All relationships were fitted using “lm” linear regression (Xiao et al., 2021).

All the green manure treatments improved maize grain yield, straw and total biomass compared to CK in both years (P < 0.05, Figure 1), with the highest yields observed in AL. Compared with WF, AL and CV significantly increased maize grain yield by 15 and 21% in 2021 and straw biomass by 44 and 51% in 2020 (P < 0.05), and OR decreased grain yield by 9% in 2021 (P > 0.05). Compared with WF, AL and CV increased total biomass by 24 and 21% on average over the 2 years (P < 0.05), respectively. Compared with the OR treatment, the AL and CV treatments had higher grain yields, straw yields and total yields, which increased by 7%−32%, 21%−27% and 13%−27% in 2020 and 2021, respectively.

Two-way ANOVA showed that year and green manure application had significant effects on grain yield, straw and total yield (P < 0.05), but there was no significant interaction effect between them (P > 0.05, Table 2).

Green manure incorporation increased the N concentration of maize grains, stems and leaves in 2020 and leaves in 2021 compared to CK (P < 0.05, Figure 2A). Both AL and CV increased the grain N concentration compared to WF, but the mean increase intensity differed between AL (23%, P < 0.05) and CV (5%, P > 0.05). OR decreased the grain N concentration in 2020 (P > 0.05) and 2021 (P < 0.05, Figure 2A). The P concentration in maize plants was lowest in CK and highest in AL (Figure 2B). The grain, stem and leaf P concentrations in the AL treatment increased by 3%, 20% and 9% on average, respectively, compared with those in the WF treatment.

Both total maize N and P uptake showed consistent changes in 2020 and 2021, with the highest uptake in AL, followed by CV and OR, and the lowest uptake in CK (Figure 3). All fertilizer treatments significantly increased the maize grain, straw, and total N and P uptake compared to CK in 2020 and 2021 (P < 0.05, Figure 3). Compared with WF, the maize grain, straw and total uptake in AL and CV increased by 49 and 21%, 87 and 88%, and 63 and 45% on average, while the RA decreased by 20%, 9%, and 11%. The AL treatment significantly increased the maize grain, straw and total N uptake compared with the WF treatment (P < 0.05, Figure 3A). AL had the highest grain, straw and total P uptake, and the straw and total P uptake was significantly higher than WF (P < 0.05, Figure 3B). Compared with CK, all the treatments significantly increased the maize grain N/P ratio except OR; AL and CV increased the leaf N/P ratio in 2020 and 2021, and OR increased the stem N/P ratio in 2021 (P < 0.05, Table 3).

Legume green manure incorporation showed an overall trend of improving N and P fertilizer use efficiencies, while non-legume green manure (OR) application showed an opposite trend in the second year (Table 4). Compared with the WF treatment, the UEN and NUE increased by an average of 63 and 103% in the AL treatment and by an average of 45 and 74% in the CV treatment, respectively; the UEP and PUE increased by an average of 35 and 66% in the AL treatment and by 22 and 41% in the CV treatment, respectively. AL significantly increased UEP and PUE compared to WF in both years (P < 0.05). In addition, the UEP and PUE in AL and CV were higher than those in OR. Two-way ANOVA showed that both N and P fertilizer use efficiencies were significantly affected by green manure type (P < 0.05), and there were no significant interactions between green manure type and year (P > 0.05, Table 4).

AL significantly increased the soil available and total N contents compared with the other treatments in 2020 and 2021 (Figures 4A, B, P < 0.05). The soil available N of CV and OR was significantly higher than that of CK and WF in 2021 (P < 0.05, Figure 4A), and there were no significant differences in total N among CK, WF, CV and OR in both years (P > 0.05, Figure 4B). The soil available P in AL and CV was significantly higher than that in CK and WF in 2020 and 2021 (P < 0.05, Figure 4C). The soil total P content had no significant changes among fertilizer treatments (P > 0.05, Figure 4D).

The SEM explained 99 and 96% of the variations in NUE and PUE, respectively (Figure 5). NUE was directly affected by grain and straw N uptake, and PUE was directly affected by grain and straw P uptake. Moreover, green manure indirectly affected NUE due to soil available N and P contents and N uptake in grain and straw, while green manure indirectly affected PUE due to soil available N and P contents and P uptake in grain and straw (Figure 5). Maize total biomass exhibited a significantly positive linear relationship to NUE and PUE (Figures 6A, B), and there were significant positive linear relationships between NUE and PUE, grain yield and grain N/P ratio in 2020 and 2021 (P < 0.05, Figures 6C–E).

In the present study, alfalfa green manure incorporation significantly increased maize grain and straw yields in 2020 and 2021, and common vetch had a positive effect on grain yield in the second year, while rape green manure reduced grain yield in 2021 (Figure 1). Previous studies showed that the application of milk vetch green manure significantly increased maize yield by 31% in a 2-year study (Tao et al., 2017), while legume hairy vetch application had no significant effect on maize yield, and non-legume annual ryegrass and cereal rye green manure application decreased maize grain yield by 4% in a 5-year study (Qin et al., 2021). A meta-analysis showed that legume and non-legume green manures increased maize yield by 12 and 9% in northern China, respectively (Ma et al., 2021), while non-legume cover crops resulted in a 4% yield decrease (Abdalla et al., 2019). These results are similar to our findings, suggesting that the effects of green manure incorporation on crop yields are closely associated with green manure type and application year.

We found that legume green manures were more effective in promoting maize growth and grain yield than no legume (Figure 1). This may be due to the following mechanisms. First, the soil N content and legume green manure N content were higher than those in the non-legume treatment (Table 1 and Figure 4A). The accumulation of N in leguminous plants is normally higher than that in non-leguminous plants because legumes can fix N₂ from the atmosphere into soils through rhizobia (Zandvakili et al., 2017), which leads to an increase in soil N supplementation and high N concentrations in leguminous plants (Dovrat et al., 2020). Second, the C/N ratio is one of the determining constraints for effective decomposition of green manure. Plant residues contain a lesser proportion of carbon to nitrogen than the 24 perfectly balanced diet soil microorganisms need, the microbes consumed the plant residues and leave the excess nitrogen in the soil. This surplus nitrogen in the soil is available for growing plants. Plant residues with a C/N ratio >24 result in a temporary nitrogen deficit (immobilization), and those with a C/N ratiol < 24 result in a temporary nitrogen surplus (mineralization) (United States Department of Agriculture, 2011). The C/N ratios of alfalfa and common vetch were much lower than that of oilseed rape in the present study (Table 1). The decomposition and mineralization rates of legume residues are faster than those of non-legume residues due to the narrow C/N ratio of legume green manures (Calegari et al., 2013; Toom et al., 2019), which release N and P faster, thus improving maize growth and yield. However, non-legume green manure with a high C/N ratio would first stabilize soil available N, which limited the N applied for crop growth (Li et al., 2020). Furthermore, green manure application may increase available P for succeeding crops (Haynes and Mokolobate, 2001), and net P mineralization from legume green manure is commonly positively correlated with the P concentration in green manure (Pypers et al., 2005). Legume green manure application could increase grain yield by improving a large amount of soil active P (Gao et al., 2016). Although the P concentration in oilseed rape was the highest in the present study, the soil available P content was lower than those in AL and CV (Figure 4) because the low N concentration and large C/N ratio of oilseed rape limit the decomposition of rape residues and further affect available P release.

Maize has a relatively high nutrient requirement for N, while the average NUE of maize is low (Abbasi et al., 2013). In this study, the NUE ranged from 14 to 48%, and the application of alfalfa and common vetch increased the NUE by 103 and 74% and decreased it by 19% under oilseed rape green manure (Table 4), which coincided with the trends in maize biomass, N uptake, and available N content between fertilization treatments (Figures 1, 3A). Similarly, Murungu et al. (2011) found that the contributions of grazing vetch (Vicia darsycarpa) and forage pea (Pisum sativum) green manures to maize N uptake were higher than that of oat green manure. Red clover significantly improved maize N availability, but oilseed radish (Raphanus sativus), oat and rye green manures had no significant effect on maize N availability (Vyn et al., 2000). Chinese milk vetch increased N uptake and use efficiency by 39%−51% (Zhu et al., 2014). Increasing crop biomass and N concentration are effective ways to increase NUE (Sinclair and Vadez, 2002). The N contributions of alfalfa and common vetch green manures to maize plants were higher than those of rape green manure, causing higher NUE in the AL and CV treatments than in the OR treatment (Figure 2A and Table 4). Legume green manure is not only used as a direct source of available N for plants but also has great potential in increasing the availability of soil N to crops and preserving N (Ashraf et al., 2004). Legume green manure increased NUE mainly by directly increasing N uptake and indirectly improving soil available N and P contents (Figure 5A), suggesting that the improvement in soil available N content may be responsible for the increased crop yields and N uptake and further enhances NUE (Zhang et al., 2020). Therefore, it is suitable for enhancing N supply and NUE in alfalfa green manure-maize rotation systems.

There are several factors affecting soil available N content, which in turn affect crop N uptake and N use efficiency after green manure return to the field. First, organic N in the soil began a rapid initial fixation or mineralization, followed by a slow linear mineralization, and the C/N ratio of green manure determined the decomposition of green manure and N mineralization in soil (Cambardella et al., 2010). In general, green manure with a small C/N ratio mineralizes faster, increasing the availability of mineral N in the soil (Figure 4A), which is the major N source for absorption by crops (Radicetti et al., 2017). Second, the lower C:N ratio of green manure resulted in more rapid N mineralization, and a higher C/N ratio of green manure can prolong the microbial fixation of N available in the intensive cropping system, and its effect is more significant in the early stage of maize growth, thus limiting the N uptake and potential yield in the short term (Radicetti et al., 2017). Moreover, legume and non-legume green manure incorporation exhibited differential responses to the soil microbial community (Khan et al., 2020). Leguminous green manure rich in N easily decomposes following application, probably increasing the soil microbial functional community and soil enzyme activities (Chavarría et al., 2016), which might be attributable to carbon availability in the early stages of maize growth, further enhancing crop nutrient absorption. This was supported by our findings that AL and CV significantly increased soil urease activity and the relative abundance of Proteobacteria (He, 2022). In addition, lower N uptake in maize directly caused the soil available N content to increase with oilseed rape application in 2021. In our study, oilseed rape green manure incorporation reduced the N uptake and NUE in the second year, mainly attributable to maize grain yield reduction (Figures 1A, 3A and Table 4), which may be due to the large C/N ratio of oilseed rape and slow decomposition rate, resulting in the fixation of organic N and little contribution to maize N supplied (Radicetti et al., 2017; Carciochi et al., 2021).

We found that alfalfa and common vetch green manure incorporation increased the P uptake of maize straw, while the effects on maize grain P uptake were insignificant (Figure 3B). Previous studies have found that the P uptake in rice straw with alfalfa and broad bean green manure application was higher than that in chemical fertilizer alone (Gao et al., 2016), and maize P uptake in cobs had no significant changes under the O. violaceus green manure-maize system compared with the continuous maize system (Zhang et al., 2022). The PUE of AL and CV increased by 35%−85% in the present study (Table 3), suggesting that short-term application of legume green manure has the potential to improve the maize PUE. Sesbania green manure application increased P uptake and the recovery efficiency of P and PUE in a rice–wheat double system (Mitran and Mani, 2017). Alfalfa and broad bean green manure in combination with chemical fertilizer significantly increased rice PUE by 10%−14% compared with the application of chemical fertilizer alone (Gao et al., 2022). Green manure incorporation can increase soil available P content and crop P uptake and then improve PUE (Pavinato et al., 2017; Gao et al., 2022). Wang et al. (2021) reported that soil properties such as total N, available N, microbial biomass C and N were closely related to PUE. In the present study, SEM analysis showed that the improvement in PUE was attributed to soil available N and P contents and maize P uptake (Figure 6B).

Phosphorus cycling-related microbial and enzyme activities transform soil P through dissolution, mineralization, and absorption, converting soil insoluble P into inorganic P, which is more easily absorbed by plants (Gao et al., 2019). Specifically, the organic instability pool in surface soil was increased through mycorrhizal colonization of green manure, and the arbuscular mycorrhizal fungal abundance of subsequent crops was increased (Arruda et al., 2021). By coincidence, maize plants preferred the mycorrhizal pathway at suboptimal soil available P (Zhang et al., 2021). In addition, organic acids released during the decomposition of green manure can enhance P availability by chelating with aluminum and iron oxides, reducing the number of binding sites and reducing soil P adsorption strength by dissolving soil mineral P (Haynes and Mokolobate, 2001). Moreover, green manure effectively increased the abundances of P-solubilizing bacteria and enhanced phosphatase activity (Wang et al., 2021).

Balanced nutrient supply is the key factor in nutrient use efficiency and increased crop production (Janssen, 1998; Dash et al., 2015). There was a positive linear correlation between NUE and PUE in our study (Figure 6C). The crop N/P ratio is a direct function of N uptake and an inverse function of P uptake, and its efficiency depends on the range of the N/P ratio. Alfalfa and common vetch green manures increased the maize grain N/P ratio, while oilseed rape had no significant effect on it compared with the winter fallow, and there was a positive relationship between grain yield and grain N/P ratio (Figure 6D), suggesting that lower C/N ratio of legume green manure could enhance maize N and P uptake and consequently increase maize yield owning to the mineralization of green manure (United States Department of Agriculture, 2011). Previous studies showed that crop P uptake and PUE were positively correlated with P concentration in green manure and negatively correlated with the C/P ratio of green manure (Garg and Bahl, 2008; Gao et al., 2022). This study showed the different result that the applicaiton of oilseed rape with the highest P concentration and the lowest C/P ratio had the lowest P uptake and PUE, as well as NUE. The probably reason is that the higher C/N ratio of oilseed rape limited the N supplied and P releasing from oilseed rape residues (Calegari et al., 2013; Toom et al., 2019). Legume green manures rich in N and P contents, in particular N content, may enhance NUE and PUE by improving soil available N and P. In general, legume green manure incorporation generally has positive contributions to N and P in cropping systems.