Root-zone fertilization improves crop yields and minimizes nitrogen loss in summer maize in China

It is urgently to minimize nitrogen (N) loss while simultaneously ensuring high yield for maize in China. A two-year field experiment was conducted to determine the effects of root-zone fertilization (RZF) and split-surface broadcasting (SSB) on grain yield, N use efficiency (NUE), and urea-15N fate under different N rates (135, 180 and 225 kg ha−1). Results showed that RZF increased grain yield by 11.5%, and the N derived from fertilizer (Ndff%) by 13.1–19.6%, compared with SSB. The percentage of residual 15N in the 0–80 cm soil was 37.2–47.4% after harvest; most 15N (64.4–67.4%) was retained in the top 20 cm. RZF significantly increased the N apparent recovery efficiency (NARE) and 15N recovery in maize by 14.3–37.8% and 21.9–30.0%, respectively; while decreased N losses by 11.2–24.2%, compared with SSB. The RZF of urea can be considered a slow-release fertilizer, which better matches maize N demand and effectively reduces N losses. Overall, RZF achieved yields as high as the SSB, but with a 20–25% reduction in N application. These results help improve our understanding of N fate in the maize cropping system, and may help guide recommendations for N management in southeastern China.

to 300 kg N ha −1 . The optimum N dose was 169 kg N ha −1 in Jianghuai area of Anhui province. Over-application of N fertilizer does not significantly increase maize yields but does result in accumulation of nitrate N (NO 3 -N) in the soil, and it causes large amounts of N to be lost to the environment 9,21 . Maize is often associated with large surpluses of soil mineral N that may reach 200 kg N ha −1 after harvest 22,23 . Therefore, it is essential to develop appropriate N management methods to reduce N losses and get a substantial and consistent yield increase of maize in this region. Such methods should not only optimize N inputs so that crops can achieve maximum yields and high NUE, but also minimize N losses to the environment.
Extensive studies have been performed to reduce N rate and enhance NUE, such as increasing the number of applications (split-surface broadcasting, SSB) according to the plant's N needs 3,24,25 , using urease inhibitors and controlled-release fertilizer 26,27 , and planting more efficient maize varieties. Each method has a cost: urease inhibitors and controlled-release fertilizer are expensive; SSB applications are labor-intensive. The costs of farm labor in China are rising in general because the workforce has decreased as farmers have aged and people have moved to industrialized areas. Therefore, both farmers and the government are demanding simplified fertilization methods, such as one-time fertilization, provided the methods do not decrease crop yields 28 . Selecting the right fertilizer rate and placement is a key role into influencing plant growth and nutrient uptake, which are vital to reduce total N input and losses 29 . Identifying the optimum fertilization rate and placement has become extremely important for one-time fertilization programs. In several recent studies, one-time urea deep placement has been reported to increase crop yield and reduce N losses [30][31][32] . In our previous work, we also found that one-time nitrogen root-zone fertilization (RZF) significantly reduced N rates and N losses without reducing yields in rice paddy fields 28 and the wheat-soil system 12 . RZF is a more exact deep placement of fertilization according to the specific crop 28,29,33,34 . For rice, fertilizer applied into 10 cm deep holes positioned 5 cm from the rice roots was the suitable RZF pattern, which increased rice yield while reduced fertilizer nitrogen loss 34 . For summer maize, we found that N applied all at one time as a basal fertilizer into a hole that was 5 cm from the seed and 12 cm deep was the effective RZF pattern for increasing nitrogen use efficiency and grain yield 33 . However, few studies have assessed the fate of isotopic urea ( 15 N) or the N losses associated with RZF in summer maize drylands. The isotopic ( 15 N) technique is widely used to evaluate the fate of fertilizer N and use efficiency in farmland ecosystems 2,3,12,22,25 . A better understanding of N fate in a summer maize cropping system would help improving N management for increasing maize yield while reducing N loss. Therefore, a 2-year consecutive field experiment was conducted to investigate the effect of RZF on maize yield, N uptake, NUE and the fate of N using the 15 N tracer technique in Anhui province, one of the main maize planting regions of China.

Results
Maize yields. Across two years, the grain yield and biomass of maize was significantly affected by both application rate and method of N fertilizer (Table 1). Grain yield and biomass were significantly higher in all N application treatments (7.70-10.07 t ha −1 and 12.91-16.69 t ha −1 , respectively) compared with the control treatment (5.47-5.68 t ha −1 and 8.92-9.69 t ha −1 ). Within the same application method (RZF or SSB), both grain yield and biomass increased with increased doses of N (from 0 to 225 kg ha −1 ). The grain yields for treatments N135 and N180 in RZF were significantly higher than SSB135 and SSB180 by 19.1% and 12.6%, respectively, in 2015, and by 15.5% and 8.1% in 2016. However, there was no significant difference in grain yield between SSB and RZF at the highest N dose (N225). Similarly, the biomass of RZF135 was 7.0% and 9.4% higher than that of SSB135, respectively, in 2015 and 2016, and 5.9% higher for RZF180 than SSB180 in 2016. Over all N doses, grain yield and biomass in RZF were 11.5% higher and 6.0% higher than in SSB, respectively. N uptake by maize. Both the N application rate and method significantly affected the N accumulation in both grain and the whole plant of maize ( increased with increasing dosage of N. The N accumulation in the grain was significantly higher (16.1% at N135, 15.8% at N180, 15.0% at N225) in RZF than in SSB in 2015, however there was no difference for N180 or N225 in 2016. Similarly, the total N accumulation for in RZF135, RZF180 and RZF225 were significantly higher than in the same dosages of SSB (by 11.3%, 10.6% and 13.5%, respectively, in 2015, and by 7.8%, 7.3% and 9.3% in 2016).
Nitrogen use efficiency. All three NUE indexes [the N apparent recovery efficiency (NARE), N agronomic efficiency (NAE), and N partial factor productivity (NPFP)] were significantly affected by fertilization method over two consecutive years. NARE and NPFP were also significantly affected by N application rate in both years, whereas NAE was not affected by the N application rate in 2016 ( Plant nitrogen derived from fertilizer and soil. The N derived from fertilizer (Ndff) in grain and whole plant were significantly affected by the method and rate of N fertilizer. As showed in Table 4, the Ndff in both grain and the whole plant increased with increasing the application dose of N. RZF significantly increased the Ndff in grain and the whole plant by 12.6-28.5% and 22.0-30.1% compared with SSB, respectively, and increased the Ndff (%) by 13.1-19.6%. However, the percentage N derived from soil (Ndfs) for RZF was significantly lower than that of SSB treatments. In general, RZF significantly increased Ndff (%) but decreased Ndfs (%), compared with SSB, and the proportion of fertilizer uptake at the highest doses (N225) was higher than at the lowest N doses (N135).   Table 2. Effects of method of application of N and its dose on mean N uptake (kg ha −1 ) by maize in 2015 and 2016. Different small letters within the same column represent significant differences (P < 0.05). CK: N application 0 kg ha −1 ; SSB: two-split surface broadcasting; RZF: one-time root-zone fertilization. ns means not significant; **Significant at P < 0.01; ***Significant at P < 0.001.  Table 3. Effects of method of application of N and its dose on three measures of N use efficiency (NUE) in 2015 and 2016. Different small letters within the same column represent significant differences (P < 0.05). CK: N application 0 kg ha −1 ; SSB: two-split surface broadcasting; RZF: one-time root-zone fertilization. ns means not significant; *Significant at P < 0.05; **Significant at P < 0.01; ***Significant at P < 0.001. Fate of urea-15 N in maize-soil system. The N application method significantly affected the N recovery in maize and potential losses, but did not affect the N residual in soil (

Discussion
RZF increased maize grain yield by an average of 11.5% compared with SSB over two consecutive years. In our previous studies, rice yield of one-time RZF of N was 19.5% greater than the yield for 3-split surface broadcasting 34 . Significant increase in crop yield for improved fertilization method have been reported in both rice and wheat compared with surface broadcasting 15,22 . Deep N application (10 cm) could increase wheat yields by 30.3% on average compared with surface broadcasting 22 . Yao et al. 15 found that one-time deep placement of urea  Table 4. Maize plant N derived from 15 N-labeled urea (Ndff) and from soils (Ndfs) in 2016. Different small letters within the same column represent significant differences (P < 0.05). CK: N application 0 kg ha −1 ; SSB: two-split surface broadcasting; RZF: one-time root-zone fertilization. ns means not significant; *Significant at P < 0.05; ***Significant at P < 0.001. increased rice yields by 10% compared with surface broadcasting. Interestingly, grain yield of RZF with low N doses (135 kg N ha −1 ) was equal to that of SSB with middle N doses (180 kg N ha −1 ), and grain yield of RZF180 was also equal to that of SSB225 in 2016, and even 11% higher than that of SSB225 in 2015. That is, when the rate of N application was reduced 20-25% in the root-zone fertilization, RZF135 and RZF180 still produced almost the same yield of maize grain as SSB180 and SSB225, respectively, in which urea was split-fertilized. Under the low and middle N doses (135 and 180 kg N ha −1 ), RZF significantly increased grain yield by 11.8-15.9% compared with SSB, whereas there was no significant difference between RZF and SSB for the high N doses (225 kg N ha −1 ) application.
Maize yields varied between 2015 and 2016, especially among the RZF treatments, which may have been driven by environmental conditions, particularly high temperature and heavy rainfall. In the 2016 maize season, the temperature (29.1 °C on average) in July was higher than that in 2015 (26.4 °C). The rainfall in July 2016 (533.8 mm) was extremely higher than that in July 2015 (259.1 mm). More importantly, the rainfall in the first 8 days of July was 487.6 mm (91% of this month), particular in July 3, the rainfall reached up to 253.2 mm (Fig. 2). Heavy rainfall would lead to serious N leaching to the deeper soil layers which cannot be absorbed by maize roots. And therefore, the grain yield of RZF was limited by the reduction in nitrogen absorption in 2016. However, half of N fertilizers for SSB were applied on July 8, 2016 (45 days after sowing). Therefore, the N absorption for plant in SSB was less inhibited by rainfall than that in RZF, and the grain yield for SSB kept in a relatively stable level during 2015 and 2016. This may explain why the available N of RZF treatments in 2016 was lower than it was in 2015. Yields in the RZF treatment were slightly higher in 2015 compared with 2016, whereas yields in the SSB treatments were not significantly different between 2015 and 2016 (Table 1). In general, one-time RZF of urea achieved higher grain yields and biomass than 2-split surface broadcasting for the summer maize system used in this study, especially in the low and middle N application rates. Our study suggests that one-time RZF can be an effective way to reduce the amount of N fertilizer, which deserves more attention and further research in the future.
The N accumulation in plant (grain and straw) was 144.3-191.4 kg ha -1 in the N application treatments, which was similar to the results in the adjacent area of this study 35,36 . Zheng et al. 36 reported that N accumulation for the whole plant was 150.1-176.5 kg ha −1 in the N application treatments in region along Huai River of Anhui province. However, RZF increased the total uptake of N by 7.2-13.5% compared with SSB treatments, indicating that RZF of urea could greatly promote fertilizer N uptake. This is consistent with previous studies by Liu  Table 5. The fate of 15 N-labeled urea in the 2016 maize growing seasons. Different small letters within the same column represent significant differences (P < 0.05). CK: N application 0 kg ha −1 ; SSB: two-split surface broadcasting; RZF: one-time root-zone fertilization. ns means not significant; **Significant at P < 0.01; ***Significant at P < 0.001. et al. 28 and Yao et al. 15 , who found that one-time RZF or urea deep placement significantly increased N uptake in rice compared with 3-split surface broadcasting treatments. Based on two-year study, the NH 4 + -N content was higher in the root zone than that of surface broadcasting during the early growth stage, which supplied continuous high levels of N for plants, and thus promoted root growth and nutrient uptake under RZF and deep placement of urea 15,28 . In contrast, when urea was broadcast onto the soil surface, it was rapidly hydrolyzed to root growth-stimulating NH 4 + and lost through NH 3 volatilization and runoff 15,32 . Rees et al. 22 also found that N uptake by the grain and the whole maize plant in point deep placement increased by 32.5% and 7.9%, respectively, compared with surface broadcasting. Urea deep placement could minimize NH 3 volatilization, denitrification and runoff compared with surface broadcasting 15,37,38 , thereby reducing N losses and increasing N uptake by plants. However, further studies on the root spatial distribution and N uptake, and their interaction on increasing maize yield under RZF are needed.
Furthermore, using the 15 N tracer technique, we monitored the uptake of N by plants to determine whether N was coming from the soil or from fertilizer. We found that the Ndff (%) in whole plant tissues averaged 18.4-28.5% compared with 71.5-81.6% of plant N derived from soil (Table 4). This is consistent with previous studies by Wang et al. 3 and Rimski-Korsakov et al. 39 , who reported that 67.6-73.2% and 54-78%, respectively, of plant N coming from the soil. Chen et al. 40 also found that N from soil organic matter contributed approximately 83% of the total N in the rice-wheat cropping system, and the proportion was about 88% when crop residues were applied. These results suggest that the mineralization of soil organic matter is an important source of N for crops. However, RZF significantly increased the Ndff (%) by 13.1-19.6% compared with SSB, while the Ndfs (kg ha −1 ) in RZF was not significantly different than that of SSB (p > 0.05) under the same N application rate ( Table 4). That is, one-time urea RZF significantly reduced nitrogen losses without depleting soil nitrogen. This finding was in accordance with Rees et al. 22 , who found that point placement significantly increased the Ndff (%) of summer maize by 27.8-33.3% compared with surface application. Recently, Yao et al. 15 also reported that one-time urea deep placement significantly increased the Ndff (%) of rice by 62% compared with surface broadcasting. Urea deep placement could maintain a significant higher NH 4 + -N content than surface broadcasting in the root zone 15,28,41 . When roots encounter a nutrient-rich zone or patch, they can enhance its proliferation and the capture capabilities of fertilizer nutrients 38,41 . In contrast, when urea was broadcast onto surface, the nitrogen nutrient was easily hydrolyzed to NH 4 + and lost through NH 3 volatilization and runoff, because the fertilizer nutrients stayed away from the roots 15,41 . Thereby, the N uptake derived from fertilizer was significant higher in RZF than that in SSB. Overall, these results indicate that RZF of urea could greatly enhance the plant N uptake from fertilizer and N use efficiency.
RZF significantly increased NARE and N recovery by 14.3-37.8% and 21.9-30.0%, respectively, compared with SSB (Tables 3 and 5). This is consistent with the findings of Rees et al. 22 , who reported that the maize recovery of fertilizer N following point placement was 38.9% higher than for surface application. Similarly, previous studies reported that deep placement of N fertilizer could increase N recovery efficiency (NRE) of rice by 26-93%, and significantly increase 15 N recovery, compared with surface broadcasting 15,28,41 . Our previous study carried out in the typical lime concretion black soil (heavy texture) also showed consistent results as the present study (silty clay) 42 . It has been reported that deep placement of urea increases growth of lateral roots at deep soil layers and also increases taproot diameter and length 43 . Deep placement of urea has also been shown to result in persistently high levels of nutrients in plant-available form close to roots 32 , thereby increasing uptake by plants. Moreover, deep-placed urea decreases NH 3 volatilization, denitrification and runoff compared surface broadcasting 15,37,38 , and thus reducing N losses and increasing N uptake by plants. Moreover, deep placement increased the NAE and NPFP of maize under the same rate of N application (Table 3), which was in agreement with the results described by Liu et al. 41 , who reported that deep placement of N significantly increased NAE and NPFP by 31-51% and 10-16%, respectively, in paddy fields compared with surface broadcasting. Recently, Pan et al. 44 also found that mechanical deep placement of N fertilizer significantly enhanced NAE by 19.5-50.4% in a rice cropping system of South China, compared with surface broadcasting. Therefore, RZF is an effective way to improve NUE for maize systems.
The potential N losses were significantly reduced in the RZF treatments. Lower potential N losses for deep placement also found by Cai et al. 38 , who reported that deep placement greatly reduced the total N loss from 42-67% to 10-27% compared with surface broadcasting. The N loss fertilizer is closely related to the application method and N rate 3,6,15 . Previous studies have reported that the N loss was significant lower with one-time deep placement than split-surface broadcasting for rice 15,28,40 . However, Chen et al. 45 reported that compared to 3-split application, unaccounted-N loss in winter wheat for one-time band deep placement was increased by 21.7%. Therefore, optimal one-time deep application may be different for different crop to decrease N loss. Lower fertilizer N loss for RZF treatments may be due to a slow-release fertilizer, which better matches the N demand of maize plants and remarkably shrinks NH 3 volatilization 15,28 . In contrast, there was no significant difference of the total residual 15 N between RZF and SSB at the same application rates of N (Table 5). At harvest, the total residual 15 N-labeled fertilizer in the 0-80 cm soil layer was 37.2-47.4% of 15 N fertilizer, with 64.4-67.4% of that retained in the 0-20 cm layers ( Fig. 1 and Table 5). These numbers are similar to those of Wang et al. 3 and Yang et al. 46 , who found that about half of total residual 15 N-labeled fertilizer retained in the 0-20 cm soil layer at maize harvest. For wheat, most of the residual N (76.8-87.0%) was retained in the 0-20 cm soil layer in southeastern China 45 . These results indicate that residual N does not move much, and the N leaching risk was low. In general, nitrogen can be lost mainly through ammonia volatilization, runoff, leaching and denitrification 15,38 . However, we just only investigated the total potential losses of the SSB and one-time RZF of urea in the present study. Therefore, more studies should be carried out to directly measure the N losses, such as ammonia volatilization, runoff, leaching and denitrification in future. Nevertheless, RZF is a method for improving the efficiency of fertilizer application and achieving low levels of leaching, while there is room to further develop and apply root-zone fertilization machinery.

Conclusions
RZF increased maize grain yields by an average of 11.5% compared with SSB, and RZF achieved high yield as SSB with a 20-25% reduction in the N dose under 135-180 kg N ha −1 . The plant N derived from fertilizer averaged 18.4-28.5% compared with 71.5-81.6% derived from soil. RZF significantly increased the Ndff (%) by 13.1-19.6% compared with SSB. The total residual 15

Materials and Methods
Experimental site. A two-year experiment was conducted in Dongzhi county (30°17′N, 117°4′E), in Anhui province, southeastern China. The region has a subtropical, humid monsoon climate. The annual mean air temperature and the average annual precipitation from 1951 to 2008 are 16.9 °C and 1554 mm, respectively. The pattern of monthly mean air temperature and precipitation during the maize growing season from 2015 to 2016 is given in Fig. 2. The soil of the experimental site is classified as a red-yellow soil. The texture of the soil is silty clay  15 N was bordered by 50 cm-high polyvinyl chloride (PVC) frame (28 cm-wide and 60 cm-long) with 45 cm inserted into the soil and 5 cm exposed on the surface to prevent runoff and lateral contamination. We used the maize cultivar 'Longping 206' , a local prevailing cultivar. Seeds were sown at a spacing of 60 cm × 28 cm (60,000 plants ha −1 ) in all treatments. The 15 N-labeled urea (46% N content, and 10.15% of 15 N abundance ratio) was provided by the Shanghai Research Institute of Chemical Industry. The non-labeled N fertilizer was urea (46% N). The P fertilizer (135 kg P 2 O 5 ha −1 , superphosphate) and K fertilizer (180 kg K 2 O ha −1 , potassium chloride) were applied to all treatments at the time of sowing. In the RZF treatments, the urea was point deep-placed all at one time as a basal fertilizer (4.89, 6.52 and 8.15 g urea plant −1 for RZF135, RZF180 and RZF225, respectively). Both the 15 N-labeled and non-labeled urea for RZF were manually placed into a 12-cm deep hole positioned 5 cm from each maize seed using a steel-pipe tool (a structure similar to an injection syringe). In the SSB treatments, 50% non-labeled or 15 N-labeled urea was broadcast by hand at sowing, and 50% at tasseling stage. P and K fertilizers were broadcast as basal fertilizers in all treatments.
Plants were watered in accordance with typical irrigation practices, and irrigation practices were the same for all treatments. Pesticide and herbicide applications were the same for all treatments. Maize was sown by hand on May 29, 2015 and May 24, 2016, and harvested on Sept. 22, 2015 andSept. 19, 2016. Plant and soil sampling and analysis. Plants in each plot were harvested close to the ground, and then were separated into straw and grain. Two plants from each plot that had been labeled 15 N urea were harvested and separated into straw, grain, and roots; the roots were collected from the 0-60 cm soil layer and washed. The dry weight was determined after drying at 70 °C to constant weight. An aliquot of each dry sample was ground to powder and passed through a 0.15-mm sieve in preparation for total N content and isotopic 15 N analyses. Soils were sampled to a depth of 80 cm; each soil sample had a radius of 20 cm and was divided into 20 cm sections. Total N of the grain, straw and roots were analyzed using the Kjeldahl method. The 15 N abundance was measured using an elemental analyzer (Costech ECS4010, Costech Analytical Technologies Inc., Valencia, USA) coupled to an isotope ratio mass spectrometer (Delta V Advantage, Thermo Fisher Scientific Inc., USA). Soil bulk density was determined after harvesting with the cutting ring method.
Calculation methods. All 15 N was expressed as atom percent excess and corrected for background abundance (at 0.366%). The N derived from fertilizer (Ndff) and N derived from soil (Ndfs) were calculated according to the following equation 3,15 : where A is the 15 N natural abundance; B is the atom percent excess of 15 N in the plant or soil; and C is the atom percent excess of 15 N in the fertilizer N.
The N fertilizer accumulation and recovery by maize were calculated according to the methods of Wang et al. 3 . Nitrogen apparent recovery efficiency (NARE), nitrogen agronomic efficiency (NAE), and nitrogen partial factor productivity (NPFP) were calculated according to the following method 16,41,44 .
1 N N where U 0 and U N represent the total N uptake (kg ha −1 ) by the grain and straw in the N 0 plot and other N-fertilized plots, respectively; Y 0 and Y N are the grain yield (kg ha −1 ) in the N 0 plot and other N-fertilized plots, respectively; and F N is the rate of applied fertilizer N (kg ha −1 ).