Introduction

Rice (Oryza sativa), as one of the most important and stable food crops especially in Asia, feeds nearly half of the world’s population (Muthayya et al. 2014). With the continuous increase in world population, rice production plays a more important role in food security (Jing et al. 2008). As the primary limited nutrient factor in agriculture, nitrogen (N) is extremely important for maintaining crop yields and improving grain qualities (Fageria and Baligar 2005; Ladha and Chakraborty 2016; Mohammed et al. 2013). However, many farmers apply N fertilizer more than necessary to ensure high rice production. Excessive N fertilizer and improper application methods not only increase input costs, but also result in low N-use efficiency, high fertilizer losses through nitrate leaching and surface runoff, and atmosphere breakup through ammonia (NH3) volatilization and nitrous oxide (N2O) emission (De Datta and Buresh 1989; Min et al. 2012; Shang et al. 2014; Vlek and Byrnes 1986; Xing and Zhu 2000). Such N losses eventually lead to nitrate pollution of groundwater, eutrophication of water systems, and rise of greenhouse gas concentrations in atmosphere. Therefore, both environmental and economic factors must be simultaneously considered to ensure efficient utilization of N fertilizers for sustainable crop production (Delin et al. 2008; Stevens et al. 2005).

Rice is the predominant crop in the Taihu Lake region of southern China, which is often grown from June to November. The subsequent wheat season starts from November to the following May. The rice–wheat rotation system is the most important and prevailing rice-based agroecosystem in this region, which has high N fertilizer inputs (240–300 kg N ha−1 for rice and 200–250 kg N ha−1 for wheat) (Zhao et al. 2015). However, the economic benefit of planting wheat is very low in winter, and leaching loss of N is higher than that during the rice season (Zhao et al. 2012; Zhu 2008). There are many labor-intensive businesses in the Taihu Lake region, which results in a labor shortage in agricultural production. Recently, fallow farmland has become an increasingly common phenomenon in this area during the winter (Yu et al. 2014).

In China, cereal grain yields increased by 65% from 1980 to 2010, but at the same period, the utilization of chemical fertilizers also increased by 512% (Zhang et al. 2011). The Taihu Lake region is one of the most important rice-producing areas in southern China. To ensure production, a large amount of N fertilizer is put into paddy fields. However, rice yields have barely increased from excessive application of fertilizer N, and fertilizer N-use efficiency has significantly decreased. Most farmers have applied topdressing fertilizer based on their experiences, regardless of the appropriate amount and optimal timing of topdressing fertilizer as well as the inherent capacity of soil to supply N in such rice production systems.

Native soil N supply capacity declines rapidly with continuous rice cultivation unless it is restored by biological nitrogen fixation (BNF) (Fageria et al. 2003). Legumes have been proved to be beneficial for stable soil fertility and crop production (Mazzoncini et al. 2011; Shafi et al. 2007). Green manure crops can offer abundant organic matter, which improves soil physical, chemical and biological properties, maintains soil nutrients and productivity, and eventually sustains crop yields (Douglas et al. 1998; Fageria et al. 2005). Substituting winter legumes for wheat not only decreases ammonia volatilization and N loss due to runoff, but also improves rice yield and soil N supply (Yu et al. 2014; Zhao et al. 2015). Moreover, growing legumes in fallow fields after harvesting main crops can increase solar energy capture and C flux into soil, and thereby supply soil macro- or micro-organisms with food (Dabney et al. 2001).

The present study was established on a long-term field trial, where field micro-plot (labeled 15N fertilizer) experiments were also conducted. The objectives of our study were to investigate, (1) What are the effects of the preceding phases (Chinese milk vetch, fallow and wheat) on rice yields, the fertilizer N-use and recovery efficiencies in rice season?; (2) how does the timing of N fertilizer application (basal, tillering and heading stages) affect rice fertilizer N use and recovery efficiencies under different rice-based cropping systems? The results of our study will play an important role in guiding comprehensive promotion and application of legumes as substitutes for wheat and fallow under rice-based cropping systems in southern China.

Materials and methods

Experimental site information

Experiments were conducted at the Changshu Agricultural Ecology Experiment Station, Jiangsu, China (120°42′E, 31°33′N). The station is located in the Taihu Lake region in the humid subtropical climate zone of southern China. The location has an average annual precipitation of 1200 mm and average temperature of 16.0 °C over the past 10 years. The soil is a Hydragric Anthrosol (IUSS Working Group WRB 2007) developed from lacustrine deposits. Right before the long-term field experiment (2007), soil samples from the top 0.15 m of surface soil contained 2.17 g kg−1 of total nitrogen (Kjeldahl method), 0.82 g kg−1 of total phosphorus (Mo–Sb colorimetry after digested by H2SO4 and HClO4), 38.2 g kg−1 of organic matter (K2Cr2O7-volumetric method), 13.1 mg kg−1 of plant available P (Mo–Sb colorimetry after NaHCO3 extraction), 174 mg kg−1 of plant available K (Flame spectrometry after NH4OAc extraction), and had a pH of 7.6 (tested in a sample containing a 1:2.5 ratio of soil to water).

Field experiments

Chinese milk vetch (Astragalus sinicus L.)-rice (CMV–R) rotation is cultivated for the long-term experiment that has been under way since 2007, which includes five N treatments: 0 kg N ha−1 (CMV–R0), 120 kg N ha−1 (CMV–R120), 180 kg N ha−1 (CMV–R180), 240 kg N ha−1 (CMV–R240) and 300 kg N ha−1 (CMV–R300). The two control treatments consist of a fallow–rice (F–R) rotation with 240 kg N ha−1 (F–R240), and a wheat–rice (W–R) rotation with 240 kg N ha−1 (W–R240). In the past, 240 kg N ha−1 was the recommended N fertilizer application rate to guarantee high rice yield in the Taihu Lake region. The plots (25.76 m2) were completely randomized in the experimental field with three replications. For rice cultivation, N fertilizer was applied based on above treatments, and both P and K fertilizers were supplied 20 and 60 kg ha−1, respectively. For wheat cultivation, N, P and K fertilizers were applied at 225, 30 and 60 kg ha−1, respectively. For Chinese milk vetch cultivation, N, P and K fertilizers were applied at 0, 30 and 60 kg ha−1, respectively. No fertilizers were applied into winter fallow plots. N fertilizer was separately applied at basal, tillering and heading stages (4:2:4) in rice season. The N fertilizers were applied on 19 June, 5 July, and 13 August in 2012, and on 22 June, 7 July and 14 August in 2013. The P fertilizer was completely applied at the basal stage. The K fertilizer was applied twice, half at the basal stage and half at the heading stage. The fertilizers applied at the basal stage were thoroughly incorporated into the soil by ploughing, and topdressings were applied through surface broadcast by hand. Approximately 40 days before transplanting rice seedlings, the above-ground parts of Chinese milk vetch were chopped into about 0.1 m pieces and incorporated into the paddy soils by ploughing. The rice seedlings were transplanted on 20 June in 2012 and 23 June in 2013. The winter crops were sowed on 8 November in 2012 and 13 November in 2013.

Micro-plot experiments

After plowing in 2012, single polyvinyl chloride (PVC) tubes were installed as micro-plots within field plots on 5 June in 2012 (except G-R0). The lower ends of these tubes were hammered vertically into the soil to a depth of 0.45 m, and the upper ends extended approximately 0.15 m above the soil surface. Twelve rice seedlings were planted in each of these PVC tube micro-plots. The mineral fertilizers application rates of micro-plots were converted according to the field experiments, but 15N-labeled urea (10.15 atom% 15N) was supplied as N fertilizer. On 9 June in 2013, we installed three PVC tubes randomly into each of the G-R180, G-R240, F–R240 and W–R240 field plots at different locations compared to 2012. 15N-labeled urea with 20.12 atom% 15N was used in 2013. Table 1 provides the specific details of applying N fertilizer in the three PVC tubes of each field plot. The topdressing periods of micro-plots were consistent with the field experiments. Consistent with the local practice, the paddy fields were continuously flooded with water to a depth of 0.03–0.05 m, and then drained midseason for 1 week, followed by intermittent irrigation until harvest. Before the winter crop season, drainage ditches were dug to protect winter crops from waterlogging damage.

Table 1 Details of N or 15N fertilizer applications in three PVC tubes per field plot in 2013
Full size table

Sample collection and measurement

Rice yield determination in the field experiment: After rice matured, it was harvested manually from each field plot. The grains were separated from rice straws using a thresher and air-dried at about 25 °C in the sunshine for 10 days. Rice yield of each plot was weighted and recorded. About 500 g grains sampled from each plot were further air-dried for 15 days and the rice yields were calculated by deducting the moisture content.

Collection of rice samples from micro-plots: the above-ground parts of the rice plants were sampled during late October 2012, then the first 0–0.2 m of surface soil was dug out and the roots were manually collected. Previous research indicated that 90% roots of a rice plant are distributed in the top 0–0.2 m soil layer in rice paddies (Zhang et al. 2003). In 2012, the plants were washed to remove soil, separated into roots, straws and grains. In 2013, washed plants were divided into roots, straws, leaves and grains. All samples were dried at 70 °C, weighed and ground to powder.

Collection of soil samples from micro-plots: According to previous results, N fertilizer infiltration (leaching) depth cannot exceed 0.4 m in the present trial area (Tian et al. 2009). Therefore, we collected soil samples from 0 to 0.1, 0.1 to 0.2, 0.2 to 0.3 and 0.3 to 0.4 m soil depths after rice harvest in 2013. Soil samples were collected using a hollow auger with scale. Each sample represented a mixture of samples taken from five locations in each micro-plot. The soil samples were air-dried in the shade and ground to pass through a 0.149-mm sieve prior to analysis. The 15N-labeled soil and plant samples were then analyzed for soil total N content and atom% 15N using a mass spectrometer (Flash EA-Delta V, Thermo Fisher, USA).

Statistical calculations and analysis

Fertilizer N-use efficiency (%) was calculated as

$$= \frac{{{\text{plant}}\,{\text{N}}\,{\text{content}} \times {\text{plant}}\,{}^{15}{\text{N}}\,{\text{atom}}\% \,{\text{excess}}}}{{{\text{ N}}\,{\text{application}}\,{\text{rate}} \times {\text{fertilizer}}\,{}^{15}{\text{N}}\,{\text{atom}}\% \,{\text{excess}}}} \times 100$$
(a)

Residual N fertilizer in soil (%) was calculated as

$$= \frac{{{\text{soil}}\,{\text{ total }}\,{\text{N content }}\, \times \,{\text{ soil}}\, \, {}^{ 1 5}{\text{N}}\,{\text{ atom}}\% \, \,{\text{excess}}}}{{{\text{ N }}\,{\text{application}}\,{\text{ rate}} \times {\text{fertilizer}}\, \, {}^{ 1 5}{\text{N}}\,{\text{ atom}}\% \,{\text{ excess }}}} \times 100$$
(b)

Fertilizer N recovery efficiency (%) was calculated as = a + b.

The data were analyzed using SPSS 20.0 (2011, IBM, New York, USA). Data are means of three repetitions. Differences between relative distribution proportion of 15N fertilizer in plant and its fate in 2013 were examined by a two-way analysis of variance (ANOVA; Tables 6, 7), which were affected by the timing of N application, rotations and their interaction. Following the one-way analysis of variance (ANOVA) for other experimental treatments, differences between means were compared using Fisher’s least significant difference (LSD) test.

Results

Rice yields under different rotation systems

Rice yields under different rotation systems and N application rates greatly increased from 2008 to 2011 except for G-R0 (Table 2). From 2011 to 2013, rice yields of CMV–R0 and CMV–R120 continued to increase, while which barely increased for CMV–R180, CMV–R240, CMV–R300 as well as W–R240 and F–R240. For the CMV–R rotation system, no N fertilizer application resulted in the lowest rice yields, and rice yield increased with increasing N fertilizer application. There were no significant differences among CMV–R180, CMV–R240 and CMV–R300, but eventually the CMV-180 achieved the highest rice yield in 2013. Moreover, the highest rice yield increased by 31.2% when comparing yield of CMV–R180 in 2013 with that of CMV–R240 in 2008.

Table 2 Annual rice yields under different rotation systems and N application rates
Full size table

Among three rotation systems with 240 kg N ha−1, rice yield under CMV–R240 was always higher than those under W–R240 and F–R240, and rice yields between W–R240 and F–R240 had no significant differences. After long-term rotations, the total soil N content already exceeded the initial value (2.17 g kg−1) (Table 3). The soil N content gradually increased with the increase of N application. The soil N content of CMV–R300 was highest. There were no significant differences between CMV–R0 and CMV–R120, or between CMV–R180 and CMV–R240. The total N content of soil under CMV–R240 and W–R240 was higher than that of F–R240.

Table 3 Soil N content from 0 to 0.1 m soil under different rotation systems and N application rates in 2013
Full size table

Fate of labeled 15N fertilizer in 2012

Fertilizer 15N-use efficiency, residual N fertilizer in soil and total N recovery efficiency all increased at first and then decreased from 120 to 300 kg N ha−1 under the CMV–R rotation (Table 4). CMV–R180 resulted in the highest fertilizer 15N-use efficiency, residual N fertilizer in soil and total N recovery efficiency, which under CMV–R300 were significantly lowest among CMV–R rotation with different N application rates.

Table 4 Fate of 15N fertilizer under different rotation systems and N application rates in 2012
Full size table

Different rotation systems with 240 kg N ha−1 had great effects on the 15N fertilizer use efficiency, residual N fertilizer in soil, and total N recovery efficiency (Table 4). The fertilizer N-use efficiencies under CMV–R240 and F–R240 exceeded that under W–R240 by 55.6 and 58.8%, and the N recovery efficiencies under CMV–R 240 and F–R240 exceeded that under W–R240 by 66.0 and 44.3%. Residual N fertilizer in soil under CMV–R240 was the highest (24.6%), and there was no difference between W–R240 and F–R240.

Comparisons of N content and 15N content in rice

Under CMV–R240, the N contents of rice roots, straws, leaves, grains and total N content were higher than those under CMV–R180, W–R240 and F–R240 in most cases (Table 5). Almost 50% N assimilated by rice plants was distributed in rice grains. The 15N content derived from fertilizer in rice under F–R240 was higher than those under CMV–R180, CMV–R240 and W–R240, and that under CMV–R180 was the lowest. The proportion of the 15N content accounted for total N content in rice was in the following order: F–R240 > W–R240 > CMV–R240 > CMV–R180.

Table 5 N content per hectare into various rice plant organs under different rotation systems and N application rates in 2013
Full size table

Relative distribution proportions of 15N fertilizer in rice under different N application periods

The fertilizer 15N distribution in different rice organs was significantly affected by the timing of N application or rotation systems, but not by the interaction between these two factors except for the straws (Table 6). There was almost twice as much 15N fertilizer in the roots or leaves during the basal and tillering stages compared to that at the heading stage. The fertilizer 15N in straws under CMV–R240 was higher than that under W–R240 or F–R240 at the basal fertilizer stage. However, more 15N fertilizer was distributed in straws under W–R240 than those under CMV–R240 and F–R240 at tillering and heading stages. The 15N fertilizer distributed in grains accounted for more than 40% of total 15N fertilizer in rice during the basal and tillering stages. Fertilizer 15N contributed relatively more to the rice grains at the heading stage than at the basal and tillering stages. Furthermore, fertilizer 15N distributed in rice grains under CMV–R240 and F–R240 was higher than that under W–R240 at the heading stage.

Table 6 Relative distribution proportion of 15N fertilizer in various rice organs under different rotation systems and N fertilizer application stages
Full size table

Fate of 15N fertilizer under different N application periods in 2013

The 15N fertilizer use efficiency, residual 15N fertilizer in soil and total 15N fertilizer recovery efficiency were significantly influenced by the timing of N application or rotation systems and the interaction between them (Table 7). Fertilizer 15N-use efficiency reached more than 50% at the heading stage, and which was twice higher than those at basal and tillering stages. Residual 15N fertilizer in soil under CMV–R240 exceeded that under F–R240 by 30.6 and 43.6% at the tillering and heading stages, respectively. It was 95.3 and 23.5% higher than that under W–R240 at the tillering and heading stages, respectively. The total 15N fertilizer recovery efficiency was lowest at the basal fertilizer stage. However, applying N fertilizer at the heading stage resulted in the highest 15N fertilizer recovery efficiency. The total 15N recovery efficiencies under W–R240 were lower than those under CMV–R240 and F–R240 at basal and tillering stages.

Table 7 Fate of 15N fertilizer applied in micro-plots as affected by timing of N application and rotation systems
Full size table

Comparisons of distribution in rice and fate of 15N fertilizer between CMV–R180 and CMV–R240 in 2013

The 15N fertilizer distributed in rice roots, straws, leaves and grains had no differences between CMV–R180 and CMV–R240 no matter which N application stage (Table 8). The 15N fertilizer distributed in rice roots and leaves was significantly higher at basal and tillering stages than at the heading stage. The 15N fertilizer was distributed in grains at the heading stage more than that at the basal and tillering stages. There were no significant differences among fertilizer 15N-use efficiency, residual 15N fertilizer in soil and total 15N fertilizer recovery efficiency between CMV–R180 and CMV–R240 at basal and tillering stages. However, at the heading stage, they were significantly higher under CMV–R180 than that under CMV–R240. The fertilizer 15N-use and recovery efficiencies reached the highest at the heading stage. Residual 15N fertilizer in soil was highest at the tillering stage.

Table 8 Relative distribution in rice organs and fate of 15N fertilizer between G-R180 and G-R240 under different N application stages
Full size table

Discussion

Using Chinese milk vetch as a cover crop to improve paddy soils, rice yield increased substantially under different N levels through a long-term incorporation of CMV, even without added N fertilizer (Table 2). The highest rice yield was achieved under a relatively lower N fertilizer application rate. A curvilinear relationship was found between the N application rates and rice yields under CMV–R rotation in different years (Fig. 1). According to the fitting equation, the optimum N application rates resulting in the highest yield were 220, 243, 234 and 220 kg N ha−1 in the year 2008, 2011, 2012 and 2013, respectively. Actually, the highest rice yield was achieved by CMV–R180 in 2011 and 2013 (Table 2). There were no significant differences among CMV–R180, CMV–R240 and CMV–R300 for rice yield. However, the fertilizer N-use efficiency, residual N fertilizer in soil and total N fertilizer recovery efficiency began to decrease under CMV–R240 and CMV–R300 (Table 4). Among the three rotation systems with 240 kg N ha−1 tested here, rice yield was always ranked in the following order: CMV–R > W–R > F–R. Many studies have confirmed the beneficial effects of winter legumes on the subsequent rice crop, such as higher rice yield, lower N fertilizer requirements and environmental risks (Dalal et al. 2011; Kim et al. 2012; Yu et al. 2014; Zhao et al. 2015; Zhu et al. 2012). Legumes fix substantial amounts of atmospheric N2, and the subsequent incorporation of legume residues into the soil replenishes soil N pools and makes a significant amount of N available to the succeeding main crops (Ibewiro et al. 2000; Mazzoncini et al. 2011; Zentner et al. 2006; Zhu et al. 2012). However, Chinese milk vetch alone was far from sufficient to satisfy N demands for high rice yield. In the present study, rice yield of CMV–R120 was significantly higher than that of CMV–R0, and even had achieved comparable yields with W–R240 and F–R240, with only half N fertilizer application rate. On the other hand, the total soil N content of CMV–R120 was similar to F–R240, and even less than W–R240 (Table 3). These results suggested the total N content of soil was an important reference of soil fertility and productivity, but not the only criteria. Based on the above results, the use of CMV should be combined with 180 kg N ha−1 to reduce N fertilizer inputs and underlying environmental risks while maintaining high rice yields.

Fig. 1
figure 1

Relationship between average rice yield and N fertilizer application rate in the year 2008 (a), 2011 (b), 2012 (c) and 2013 (d). The vertical bars represent standard deviation

Full size image

The wheat–rice double-cropping system is a typical modern intensive system with high input and high output, but also with very high resource and environmental costs due to the large amounts of mineral N fertilizers inputs during the past three decades (Cui et al. 2014; Xie et al. 2017). This rotation system resulted in high background level of N in the soil (Table 3). Furthermore, continuous cultivation and fertilization have caused many problems in soils, such as soil acidification, destruction of soil structure, reduced soil quality and productivity, as well as harm to the population and activities of soil microbes (Guo et al. 2010; Liu et al. 2010, 2013a, b). Such processes are severely damaging the ecological functions and sustainability of soils and limiting their ability to self-repair. Understandably, the fertilizer N-use and recovery efficiencies under the W–R240 rotation were much lower than that under the CMV–R240 and F–R240 rotations (Table 4). The fertilizer N-use and recovery efficiencies had no differences between CMV–R240 and F–R240, but the rice yield under CMV–R240 was higher than that under F–R240 (Tables 2, 4). According to the results of fertilizer 15N content in rice, N content under F–R rotation was derived from fertilizer N more than CMV–R rotations (Table 5). The application of organic matter enhances the amount of microbial biomass N as well as the supply of soil-derived mineralized N to the plant (Nishio and Oka 2003). Thus, the rice yield of F–R240 may depend more on the mineral N fertilizer. These results suggest that high rice yield depended not only on fertilizer N-use efficiency, but could be affected by soil structure, soil organic matter and soil microbe populations and activities as well as interactions between native soil N and fertilizer N. Furthermore, the residual N fertilizer in soil under CMV–R rotation was higher than that under W–R and F–R rotations (Table 4). One possible reason is that soil organic N sources can be assimilated by plant more than inorganic N fertilizer (Motior et al. 2011). Incorporation of legume residues provides large amount of organic N sources into soil, resulting in N fertilizer retention in the soil microbial biomass, and thus the residual N fertilizer in soil was higher under CMV–R rotation system (Rahman et al. 2009).

Based on the results in 2013, the fertilizer N recovery efficiency was highest at basal fertilizer stage (Table 7). The low N fertilizer recovery efficiency could be attributable to two factors in the present study. On the one hand, rice roots are not yet fully developed during the seedling growth period and have relatively weak capacity to uptake N and other nutrients. On the other hand, excessive N fertilizer applied at early stages of rice growth would increase the proportion of ineffective tillers. The subsequent death of ineffective tillers may waste a large amount of fertilizer N (Ding et al. 2004; Jiang and Cao 2001; Qiao et al. 2013). The fertilizer N-use efficiency and total N recovery efficiency were highest at the heading stage (Table 7). Many studies have shown that N fertilizers applied at the heading stage are most effective for increasing N uptake and grain yield (Ghaley 2012; Ling et al. 2007; Zhang et al. 2013). The N fertilizer applied at the heading stage was absorbed into the grain more than that at the basal and tillering stages. These results suggest that the N fertilizer application rate can be increased at heading stage and decreased at the early growth stage of rice, which would be very important for improving fertilizer N-use efficiency and increasing N uptake. The fertilizer N-use efficiency, residual N fertilizer in soil and total N fertilizer recovery efficiency had no differences between CMV–R180 and CMV–R240 at the basal and tillering stages, and which of CMV–R180 performed even better than CMV–R240 at the heading stage (Table 8).

Conclusions

Rice yield greatly increased through long-term CMV–R rotation and reached highest at a relatively lower N fertilizer application rate (CMV–R180). Among different rotation systems with 240 kg N ha−1, the rice yield of CMV–R rotation was always higher than F–R and W–R rotations. The fertilizer N-use efficiency between CMV–R240 and F–R240 had no significant differences, but both were higher than W–R240. Further studies suggested topdressing N fertilizer at the heading stage was the most effective for increasing fertilizer N-use efficiency and absorption by rice grains. Nevertheless, the loss rate of the basal N fertilizer was highest among three N application stages. The present study showed that CMV–R rotation can be recommended as an environmental-friendly and sustainable rice-based cropping system for southern China. Further long-term studies should focus on exploring lower N fertilizer input through optimizing the application rates of N fertilizer at basal, tillering and heading stages under CMV–R rotation system.