The experiment was conducted at the experimental field of South China Agricultural University (23˚14′N and 113˚38′E), Guangzhou, Guangdong Province, China. The experiment location was a typical site of a Chinese region of rice and fallow (or green manure) cropping systems. Soil chemical properties (0-20 cm) were organic matter of 20.4 g/kg, total nitrogen of 0.82 g/kg, nitrate nitrogen of 102 mg/kg, available phosphorus of 12.1 mg/kg, and pH of 5.33 (three years mean values). The total rainfall during the winter crop growing periods in 2016-2017, 2017-2018 and 2018-2019 were 231, 286, and 388 mm, respectively (Figure 1). The average air temperature during the winter crop growing periods in 2016-2017, 2017-2018, and 2018-2019 were 17.3, 17.0, and 18.1°C, respectively (Figure 1).

The field trials were carried out from Nov. 2016 to Mar. 2017, Nov. 2017 to Mar. 2018 and Nov. 2018 to Mar. 2019. The experiment included tillage methods of NT and conventional tillage (CT) and maturity stages of the flowering stage and milk stage. Late rice was harvested in Nov. and then wheat was manually sown. For NT, the seeds were directly sown onto the soil surface, without tillage or action to improve the soil density. For CT, the soil was plowed with a pneumatic row direct drilling machine equipped with double disc furrow openers (XG788ZK, Guangzhou Agricultural Machinery Equipment, Guangzhou, China), and wheat seeds were manually scattered on the ground. Each plot was 3 m × 4 m, with an interval of 0.5 m between plots, and each treatment was in triplicate (randomly allocated). The wheat (Shimai-1) was sown at 285 kg/ha (NT and CT). No insecticides and fungicides were used during the experiments. WCW was not irrigated during the winter crop growing periods. The compound fertilizer (N: P₂O₅: K₂O=15:6:8) was split-applied at 150 kg/ha. Forty percent was applied as a basal fertilizer, and the remaining fertilizer was applied at the jointing stage. The phenology of wheat was monitored according to the decimal code for the growth stages of cereals (Zadoks et al., 1974). Table 1 shows the detailed field management and growth period of wheat. To fully utilize soil water, the wheat in 2016-2017 were planted on the third day after the harvest of late rice. However, in 2017-2018 and 2018-2019, the wheat was planted on the ninth and eleventh day after the harvest of late rice, when the soil water content was relatively low. Compared to 2016-2017, the emergence time of wheat in 2017-2018 and 2018-2019 was delayed by 5 and 8 days, respectively. In addition, due to differences in soil management methods, there were differences in the utilization of soil moisture between NT and CT, resulting in a lower seedling emergence rate for NT compared to CT (Table 1).

The WCW was harvested when plants were at the flowering stage (DC65-69), and milk stage (DC 75-77, Zadoks et al., 1974). The fresh matter yield of WCW was determined with a 1 m² (1.0 m × 1.0 m) quadrat, and the DM yield was calculated based on DM concentration and fresh matter yield. At harvest, 5.0 kg of fresh material was randomly selected (5 cm from the ground, mowed with a sickle) from each plot for the analyses of the number of epiphytic microbes, chemical composition, and silage making. Fresh materials were choped to 20 to 30 mm length, and mixed thoroughly after transporting to the laboratory. Then 300 g of materials were stuffed into bags (polyethylene plastic, 300 mm × 400 mm × 0.2 mm, Mingkang Packing Co. Ltd., Zhongshan, China), which were degassed and sealed using a vacuum sealing machine (Sinbo Vacuum Sealer, Hong Tai Home Electrical Appliance Co., Ltd., Hong Kong, China). All the bags were put in a laboratory cabinet (18 to 29°C, dark environment) for 60 days.

Lactic acid bacteria were cultured under anaerobic conditions at 37°C for 2 d (de Man-Rogosa-sharp agar), aerobic bacteria (nutrient agar), yeasts and molds (potato dextrose agar) were cultured under aerobic conditions at 37°C for 3 d. The DM concentration was measured at 70°C for 48 hours in an oven with forced air circulation (Blast drying oven 9003A, Shandong Jinan Haineng Instrument Co., Ltd., China Co., Ltd). CP and NH₃-N concentrations were determined by Kjeldahl nitrogen determination (Nitrogen analyzer KN680, Shandong Jinan Alva Instrument Co., Ltd., Jinan, China) and ultraviolet spectrophotometer (Brodeick and Kang, 1980), respectively. The analyses of ether extract and crude fiber concentrations were conducted as outlined in AOAC (1990). Crude ash concentration was measured by burning in a muffle furnace at 550°C for 5 hours. Crude fiber (CF), neutral detergent fiber (NDF), and acid detergent fiber (ADF) concentrations were determined by a filter bag method (Ankom Technology, Macedon, NY) (Van Soest et al., 1991). The buffering capacity and WSC concentration were measured using the method described by Playne and McDonald (1966) and Murphy (1958), respectively. Hemicellulose concentration was calculated from the difference between NDF and ADF concentrations. Relative feed value (RFV) was calculated based on ADF and NDF (Rohweder et al., 1978).

The pH of the extract was measured using a pH meter (LE438 pH meter, Mettler Toldeo, Shanghai, China). The concentration of organic acid (lactic acid, acetic acid, propionic acid, and butyric acid) was measured by the high performance liquid chromatography (column: Sodex RS Pak KC-811, Showa Denko K.K., Kawasaki, Japan; detector: DAD, 210 nm, SPD-20A, Shimadzu Co., Ltd, Kyoto, Japan; eluent: 3 mmol/L HClO₄, 1.0 ml/min; temperature: 60°C).

The ANOVA method was used to assess the effects and their interactions of years, tillage methods, and maturity stages on the yield, nutritive composition, and silage fermentation quality. When ANOVA indicated a significant F-value (at the 5% probability level), multiple comparisons were performed by Duncan’s multiple range method (SPSS 17.0 for Windows; SPSS Inc., Chicago, IL, USA). The model used was

where Y was the response variable, μ was the overall mean, T_i was the effect of treatment, and ϵ_ij was the residual error.

Relationships among yield and nutritional composition or among nutritional composition and silage fermentation quality of forage were examined by Pearson correlation. Analyses for the yield, chemical composition and silage fermentation quality of different years, tillage methods, and maturity stages were carried out using the principal component analysis function. Based on the information of WCW yield, nutritive composition, and silage fermentation quality, the fuzzy membership function analysis method was used to sort the years, tillage methods, and maturity stages (OriginPro^®2021b software, OriginLab Corp., WA, USA).

Table 1 indicates the harvesting time occurred later in growth stage in 2016-2017 than in 2017-2018 and 2018-2019. As shown in Table 2, the DM and CP yields were affected by three sources of variation: years, maturity stages, and year × maturity stage interaction (P< 0.05). Although tillage methods showed no significant effects on the yield, plant height, and stem number per plant of WCW (P > 0.05), compared to CT, the DM and CP yields of NT decreased by 0.74 t/ha and 0.13 t/ha. The DM and CP yields of WCW in 2016-2017 (6.75 t/ha and 0.71 t/ha) were higher than those in 2017-2018 (3.19 t/ha and 0.43 t/ha) and 2018-2019 (2.73 t/ha and 0.34 t/ha) (P< 0.05). The WCW at the milk stage had greater DM (5.25 t/ha) and CP (0.60 t/ha) yields than that at the flowering stage (3.19 t/ha and 0.39 t/ha) (P< 0.05).

As shown in Table 3, the WCW at the milk stage had greater DM, NFE, HC, and WSC concentrations than that at the flowering stage, while the concentrations of CP, CF, ether extract, crude ash, NDF and ADF was lower (P< 0.05). Compared with at the milk stage, the WSC concentration at the flowering stage decreased 47.7 g/kg DM, and the NDF and ADF concentrations increased by 41 g/kg DM and 81 g/kg DM, respectively. Therefore, WCW had higher RFV at the milk stage (114) than at the flowering stage (96.3) (P< 0.05) (Table 2). Tillage methods showed no significant effects on nutritive compositions (P > 0.05), but NT tended to reduce NDF and ADF concentrations. The CP and ether extract concentrations in 2017-2018 (140 g/kg DM and 40.9 g/kg DM) and 2018-2019 (126 g/kg DM and 42.1 g/kg DM) were greater than those in 2016-2017 (101 g/kg DM and 22.1 g/kg DM) (P< 0.05).

The tillage methods had no significant effects on the buffering capacity, pH, and numbers of all microorganisms (P > 0.05) (Table 4). The pH and numbers of all microorganisms were significantly affected by years (P< 0.05). The highest numbers of aerobic bacteria (7.28 lg cfu/g FM) and yeasts (3.48 lg cfu/g FM), and the lowest number of lactic acid bacteria (2.00 lg cfu/g FM) presented on WCW in 2016-2017. The WCW at the milk stage had less aerobic bacteria (6.35 lg cfu/g FM) and yeasts (2.50 lg cfu/g FM) than that at the flowering stage (7.01 lg cfu/g FM and 3.28 lg cfu/g FM) (P< 0.05) (Table 4).

Tillage methods, maturity stages, and their interaction showed no significant effects on the pH value and acetic acid, propionic acid, and butyric acid concentrations of WCW silage (P > 0.05) (Table 5). Experimental years had significant effects on the pH value, lactic acid, and NH₃-N concentration of WCW silage (P< 0.05). The lactic acid concentration of silage in 2016-2017 (31.6 g/kg DM) were greater than those in 2017-2018 (20.1 g/kg DM) and 2018-2019 (21.4 g/kg DM) (P< 0.05), and the lowest NH₃-N concentration was achieved in 2016-2017 (107 g/kg TN) (P< 0.05). The lactic acid concentration of NT (17.1 g/kg DM) was lower than that of CT (26.6 g/kg DM) (P< 0.05). WCW silage at milk stage had lower NH₃-N concentration (125 g/kg TN) than that at flowering stage (169 g/kg TN) (P< 0.05) (Table 5).

Correlation analyses showed that the DM yield of WCW was negatively correlated with CP, ether extract, crude ash and NDF concentrations (P< 0.05) (Table 6). The number of aerobic bacteria was significantly positively correlated with the pH of WCW silage (P< 0.05). The concentrations of DM and WSC and number of yeasts and lactic acid bacteria of WCW prior to ensiling had a negative correlation to NH₃-N concentration of WCW silage, while the CP, crude fiber, ether extract, crude ash, ADF, and NDF concentrations of WCW, before ensiling, had a positive correlation to the NH₃-N concentration of WCW silage (P< 0.05). The buffering capacity of WCW before ensiling was negatively correlated with the lactic acid concentration of WCW silage (P< 0.01) (Table 6).

The interrelationships of yields and quality indicators of WCW were relatively consistent and steady regardless of tillage methods, while silage fermentation quality indicators varied between maturity stages (Tables 2, 3, 5; Figure 2). Twenty-seven initial variables were reduced to four variables that contained information about most of the original variables (four indicators were not related to each other) (not listed). The eigenvalues of these four principal components were all greater than 1, indicating the principle of extraction as principal components was reasonable. In the initial eigenvalues, the cumulative variance contribution rate of the four principal components extracted and loaded was 96.5%. Specifically, the WCW at the milk stage had the highest yield, DM, and WSC content (Tables 2, 3), which was clearly separated from the flowering stage along the first axis (Figure 2). On the other hand, 2016-2017 had the highest yield, WSC and NFE contents, which was clearly separated from the other two years along the first axis (Figure 2). However, there were no differences in most yield, nutrients, and silage fermentation quality of two tillage methods. This was also the reason that the 95% confidence intervals overlap. According to the results of the membership function analysis, 2017-2018 scored the highest in years, NT scored the highest in tillage methods, milk stage scored the highest in maturity stages (Supplementary Table 1).

The DM and CP yields in 2016-2017 were greater than those in 2017-2018 and 2018-2019 (P< 0.05). The growth time of WCW in 2016-2017 was longer than that in 2017-2018 and 2018-2019 in both flowering and milk stages (Table 1), which was conducive to the accumulation of more photosynthetic products and increased the yield of WCW. Additionally, in Nov. and Dec. of 2016-2017, there was higher rainfall (64 mm and 14 mm) combined with higher temperatures (19.6°C and 16.7°C), which was beneficial for the accumulation of biomass in wheat early-stage. Several studies have found that climate has a greater impact on the yield and nutritive composition of forage crops (Ergon et al., 2018; Druille et al., 2020). However, the differences in temperature and rainfall between two years in this study, and wheat had similar morphological and anatomical at harvest (2017-2018 and 2018-2019). Principal component analysis also demonstrated a high overlap between 2017-2018 and 2018-2019 wheat in terms of yield, nutrient content and information on silage fermentation quality (Figure 2A).

Notably, moisture and temperature were the two most important factors affecting the growth process of crops. In this study, due to the perennial planting of early and late rice in the region, the shallow soil water distribution, which had a positive effect on the growth of the winter crop (low rainfall in winter). In addition, the sufficient light and heat were available for crop growth in winter. The above factors were all conducive to the growth of wheat. To pursue a higher DM yield, WCW was harvested after the dough stage (Filya, 2003). However, this is not allowed when the crop rotation system is more complex, as it might cause a delay in the sowing time of the subsequent crops, which in turn reduces the economic profitability of the rotation system. It was noted that the yield of wheat in this study was lower than those of some published data (Filya, 2003; Xie et al., 2012). Except for genotype, the reasons for these results were mainly related to the different growth stage and planting environment of wheat.

In the previous study, we found that the population density of wheat by NT (1.54 × 10⁶ plant/ha) was significantly lower than that of CT (2.09 × 10⁶ plant/ha), and similar results were obtained in the experiment of no-tillage Italian ryegrass (Lolium multiflorum) (Xu et al., 2021a). However, there were no differences in the DM and CP yields between NT and CT in this study. Furthermore, there were no differences in plant height and stem number per plant between NT and CT (Table 2). This shows that the per plant dry matter and crude protein accumulation rates of NT were higher than that of CT, which can explain the consistency of their yields. Light interception has been proved to increase with the increase of population density, while the effective radiation of photosynthesis penetrating the lower vegetation layer decreases (Yang et al., 2019). This also explains that the stem number per plant of NT was slightly more than that of CT (Table 2). Except for climatic factors, soil nutrients and moisture were considered to be the main factors causing variation in the yield of crop, and proper soil structure was essential to improve soil fertility and water concentration (Xu et al., 2021b). A global meta-analysis showed that no differences in the yield of crops (rice, maize, barley, and soybean) were observed between NT and CT (Shakoor et al., 2021). In the previous study, we found that no-tillage reduced disturbance and preserved more nutrients and water in soil than CT (Xu et al., 2021b). Furthermore, under NT conditions, soil water evaporation was limited and more water was available to the crop. In this study, the key factor affecting the growth of WCW was rainfall in different years. Therefore, we speculate that NT treatment suppressed the evaporation of soil water and ensured that wheat had an adequate water supply during the filling period. It was worth noting that there were differences in water supply between tillage treatments (Xu et al., 2021b). This also compensated for the risk of yield reduction due to lower population density. The reason maybe that the seeds were directly sown on the surface of the soil, resulting in a low germination rate in this study (small contact area between seed and soil, weak water absorption of seed, and no irrigation). The sample size (3 plots replicates of 1 × 1 m²) was supposed to alter the results. Principal component analysis also demonstrated a high overlap between NT and CT wheat in terms of yield, nutrient content and information on silage fermentation quality (Figure 2C). However, the results based on the analysis of the affiliation function showed that the years and maturity stages had a great impact on forage yield and nutrition (Supplementary Table 1).

The greater DM and CP yields at the milk stage than at the flowering stage were due to the longer development cycle of WCW (Tables 1, 2), accumulating greater photosynthetic products. This was consistent with the results of some studies about delayed grass harvest (Xie et al., 2012). Compared to flowering stage, the DM and CP yields at the milk stage increased by 1.35 t/ha and 0.21 t/ha, respectively. Evidently, WCW accumulated more photosynthetic products at milk stage.

The WCW at milk stage had lowter NDF and ADF concentrations than at flowering stage (P< 0.05). Many research results showed that high concentrations of NDF and ADF had a negative impact on forage digestive effectiveness (Monteiro et al., 2021). The population density of CT was high, intensifying the competition for nutrients among plants. Several studies had also provided similar conclusions (Obour et al., 2019). It was worth noting that tillage methods showed no significant effects on the nutritive composition of WCW (P > 0.05). The sowing methods does not affect the nutrition of WCW through density and soil characteristics. It suggests that sowing methods does not affect nutritive composition of WCW through population density and soil properties. Similar results were obtained in the experiments of maize (Wang et al., 2020) and soybean (Yusuf et al., 1999). With increasing maturity, the ratio of grains to whole plants increased. This explains the higher DM concentration of WCW at flowering stage.

Previous studies have shown that the growth stage (or harvest time) was an important factor affecting the chemical composition of grasses (Stirling et al., 2022). In the early stages of growth, the leaves have a high proportion of the whole plant in grasses. Therefore, grasses have a high CP concentration. In this study, the NDF and ADF concentrations of WCW decreased by 6.7% and 24.3% from the flowering to the milk stage, respectively, whereas WSC concentration increased by 50.6%. Similar trends had been observed in some studies (Xie et al., 2012). WCW at the milk stage had lower CP concentration than that at the flowering stage (P< 0.05). The reason was that protein synthesis was weakened (dilution effect) as the growth period was delayed. Conversely, starch accumulation offsets lower NDF digestibility. Unfortunately, we did not study the WCW at the dough stage.

Before ensiling, there were many kinds of microorganisms on the surface of grasses, which would produce different metabolites during the ensiling and in turn determine the fermentation quality of silage. The number of aerobic bacteria was much higher than those of other microorganisms in this study. The number of lactic acid bacteria was not affected by the maturity stages (Table 4).

Ensiling was an effective way to preserve the nutritional quality of grasses and was a global practice to preserve the moist forage crops when drying forage is difficult (especially in the rainy season). Moreover, the general practice was to wilt the crop before ensiling, if water content is high at harvest. Years had a large influence on the silage fermentation quality of WCW in this study, while the effects of tillage methods and maturity stages on the silage fermentation quality of WCW were very limited (Table 5). This was a key issue because population density was related to the epiphytic microbial population (Table 1) and could also affect the fermentation profile. In this study, WCW at the milk stage had a lower NH₃-N concentration than that at the flowering stage. This was because the higher WSC concentration of WCW at the milk stage provided more substrate for the rapid propagation of lactic acid bacteria. Moreover, the rapid reduction of pH in anaerobic environment inhibited plant and microbial protease activities, subsequently inhibiting protein degradation. Several studies had also provided similar conclusions (Xie et al., 2012). Thus, WSC concentration was negatively correlated to NH₃-N concentration (Table 6), and it was adopted as a positive indicator in the evaluation of forage feeding potential. Although we did not study the microbial species in silage, we speculations there was substantial evidences of clostridial activity throughout. Across all silages, butyric acid content was >1.0% DM, while NH₃-N accounted for roughly 15% of total N. Also, the propionic acid concentration was very high (about 2.5%) in all the treatments. All of these responses indicate significant clostridial activity.