Mechanized transplanting with side deep fertilization increases yield and nitrogen use efficiency of rice in Eastern China

The application of nitrogen (N) fertilizer deep in soil at the same time as mechanical transplanting of rice seedlings is an effective alternative to conventional broadcasting of fertilizer, but its effects on yields and profitability have not been analysed in detail. Here, we analysed the effects of a side deep application of N fertilizer at transplanting on the N uptake, N use efficiency (NUE), grain yield, and economic profitability of two rice (Oryza sativa L.) cultivars; Yongxian15 (early season) and Yongyou1540 (middle/late season). In the field experiments, two types of N fertilizer (urea (U) and controlled-release urea (CRU)) were surface broadcasted manually (B) or mechanically fertilized at 5.5 ± 0.5 cm soil depth (M) (UB, UM, and CRUM treatments, respectively). The blank control had no N fertilizer (N0). Each N-fertilizer treatment had similar effects on N uptake, grain yield, NUE, and economic profitability in the early, middle, and late seasons. Compared with manually applied fertilizer, mechanically applied fertilizer increased grain yield and NUE in both cultivars. In Yongxian15 and Yongyou1540, the mechanical side deep application of N-fertilizer increased the N recovery efficiency by 62.50–91.57% and 24.38–64.24%, respectively, the N agronomy efficiency by 33.65–63.14% and 22.64–44.70%, respectively; and the grain yield by 6.30–11.64% and 6.23–13.11%, respectively. The CRUM treatments had the highest benefit–cost ratio because of high gross returns and low fertilization costs. The mechanized side deep application of N fertilizer can increase the efficiency and profitability of rice production.

Leaf area index (LAI) and net photosynthesis (Pn). The LAI showed similar trends among all the treatments. The treatments could be ranked, from highest LAI to lowest, as follows: CRUM > UM > UB > N0 (Fig. 2). At the panicle initiation, heading, and maturity stages, there were no significant differences in LAI between UB and UM in all three seasons. At the heading stage, the highest LAI was in CRUM in all three seasons. Pn of Yongxian15 and Yongyou1540 at the heading stage was highest in CRUM, followed by UM and UB (Fig. 3). The treatments could be ranked, from highest Pn to lowest, as follows: CRUM > UM > UB > N0. The Pn of Yongxian15 in the early season and Yongyou1540 in the middle and late seasons differed significantly between UB and UM.
Total above-ground biomass. The total above-ground biomass was affected by N application methods and N-fertilizer types (Fig. 4). Total above-ground biomass was lower in the early and late seasons than in the middle season. At the panicle initiation stage, the total above-ground biomass of Yongxian15 and Yongyou1540 was lowest and highest in N0 and CRUM, respectively; and slightly lower in UM and UB than that in CRUM. At the heading stage, the total above-ground biomass of Yongxian15 in the early season and Yongyou1540 in the middle and late seasons was lowest in N0, highest in CRUM; and significantly different between UB and UM. At the maturity stage, total above-ground biomass was highest in CRUM. The treatments could be ranked, from highest total above-ground biomass to lowest, as follows: CRUM > UM > UB > N0. www.nature.com/scientificreports www.nature.com/scientificreports/ Total N uptake. The total N uptake at the panicle initiation, heading and maturity stages are shown in Fig. 5.
The highest total N uptake of Yongxian15 and Yongyou1540 was in CRUM at all growth stages, followed by UM and UB. The treatments could be ranked, from highest total N uptake at all growth stages to lowest, as follows: CRUM > UM > UB > N0. In addition, at the maturity stage, more than 60.88% of N was allocated in the panicles, followed by the stem-sheaths and leaves.   Values are means ± SE of three plots (three replicates). Different letters on columns indicate significant difference (p < 0.05, LSD multiple test).

Ammonium (NH
www.nature.com/scientificreports www.nature.com/scientificreports/ Nitrogen use efficiency. The N dry matter production efficiency (NDMPE), N grain production efficiency (NGPE), N recovery efficiency (NRE), N agronomic use efficiency (NAE), and N physiological efficiency (NPE) of Yongxian15 and Yongyou1540 were affected by N application methods and N fertilizer types in all three seasons ( Table 2). The NDMPE of both cultivars was lowest in CRUM, followed by UM and UB. However, there was no significant difference in NDMPE between UM and UB in the middle season. The NGPE differed significantly between N0 and CRUM in the early, middle and late seasons, but not between UM and CRUM in the early and late seasons. The NRE and NAE were highest in CRUM, followed by UM and UB. The NRE and NAE differed significantly between UB and UM (P < 0.05) in all seasons. The NPE in all seasons did not differ significantly between UM and CRUM. In the middle and late seasons, the treatments could be ranked, from highest NPE to lowest, as follows: UB > UM > CRUM.
Economic profitability. There were significant differences (P < 0.05) among the all treatments in fertilizer cost, gross return, net income and benefit-cost ratio (BCR) ( Table 3). The fertilizer costs in all three seasons were significantly lower for UB and UM than for CRUM. The highest grain yield was in CRUM, and significantly higher than those in UB and UM. The lowest gross return was in N0, and significantly lower than those in UB, UM and CRUM. In the early, middle and late seasons, the treatments could be ranked, from highest net income and BCR to lowest, as follows: CRUM > UM > UB. In summary, these results demonstrated that manual surface broadcasting of urea is not an appropriate and/or an economical fertilization technology for mechanically transplanted rice.  www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
The grain yield and biomass at maturity stage of Yongyou1540 in the middle and late seasons than Yongxian15 in the early season were much higher because of the longer growth stage and larger potential yields. However, the treatments had similar effects in the three seasons (Table 1 and Fig. 4). Grain yield was much higher in UM and CRUM than in UB in the three seasons. The side deep applicator used in our study (Fig. 1a) opened a slit, uniformly placed N fertilizer at 5.5 ± 0.5 cm soil depth (the anaerobic zone), and then immediately covered it with soil ( Fig. 1c). This efficiently provided a sustained N supply to improve growth and photosynthesis characteristics of rice plants for the whole growth period 38,39 because more + NH 4 -N from urea was retained in soil for a longer time 36,39,40 , thus improving the N uptake and grain yield of the rice plants. The N-fertilizer treatments had significant effects on the number of productive panicles, grain number per panicle, total above-ground biomass and yield. Compared with UB, UM increased yield in the early, middle, and late seasons by 6.30%, 6.23%, and 7.94%, respectively, because of the higher number of productive panicles and grain number per panicle, consistent with the results of previous studies 14,37,40 .
Our results also showed that the grain yield was higher in CRUM than in UM in the three seasons. Compared with UM, CRUM resulted in a much higher grain number per panicle in the early season and much higher number of productive panicles in the middle and late seasons. This finding was not consistent with those of a previous www.nature.com/scientificreports www.nature.com/scientificreports/ study 41 . The mechanical placement of U and CRU at 5.5 ± 0.5 cm soil depth beside seedlings would inevitably create a special root microenvironment zone because of the longer persistence of + NH 4 -N in the soil 31,42,43 . Compared to deep placement of U, CRU did not significantly increase the number of productive panicles, because multiple seedlings in each hole produced the rapid tiller growth during the early growth stage of the densely  www.nature.com/scientificreports www.nature.com/scientificreports/ distributed seedlings, and the N demands of rice after the heading stage were fully met by the continuous supply of N from CRU, resulting in increased grain number per panicle in the early season.
Our results also demonstrated that total above-ground biomass during the whole growth stage and LAI at heading stage and maturity stage, and the flag leaf Pn at the heading stage, were much higher in CRUM than in UB (Figs 2, 3 and 4), consistent with previously reported results 37 . The increase in total above-ground biomass in UM and CRUM was mainly related to grain, while these treatments did not have a higher harvest index than that in UB. In a previous study, CRU was shown to promote rice root growth, Pn, and dry matter accumulation, prolong the leaf functional period, postpone leaf senescence, and enhance grain yield 20,27,28,41 . Nitrate and nitrite N concentrations at the 0-25 cm soil depth in N application treatments were much higher at panicle initiation in the early season than in the middle and late seasons, and their trends among treatments were almost the same in the three seasons (Fig. 6). At the panicle initiation, heading, and maturity stages, + NH 4 -N concentrations in soil were higher in UM and CRUM than in UB. The − − NO 3 N concentrations in soil at the heading and maturity stages were much lower in UB and CRUM than in UM, indicating that less N was lost by leaching after deep application of CRU 44 . The + NH 4 -N content in soil is considered to be the most important parameter that determines the N uptake of rice, and closely related to NH 3 volatilization from standing water (more specifically + NH 4 -N in top surface soil), according to previous studies 31,43,44 . The relatively lower concentration of + NH 4 -N in soil was enough to satisfy the N demands of rice plant growth before rapid tillering, but a large amount was absorbed during the active tillering stage and lost through NH 3 volatilization. Ultimately, CRUM was better in terms of energy saving and the environment. The After U mechanically placed deep in the soil or manually broadcasted into surface water in paddy soil, it is quickly dissolved and hydrolysed into + NH 4 -N by ureases. However, the dissolution rate of CRU is slower because the nutrients are coated with special biological materials with uniform small holes. The urease activity was lower in UM and CRUM than in UB through the whole late growth period, which prolonged N availability to rice roots. These results, which will be reported in detail in another study, are consistent with those of previous studies 42  The root length of rice seedlings at transplanting usually 5-8 cm, and the roots extend throughout the 15-20 cm soil layer by about 30 days after transplanting. The added N from urea applied onto the soil surface migrates about 7 cm 45 . By inference, N in our experimental field would rapidly extend 12.5 cm (5.5 + 7 cm) vertically and horizontally in 30 days. In the 15 days after transplanting, rice plants need a little N, and so less N should be allocated at this stage 46 ; in the middle growth stage, a large amount of N nutrition is supplied to rice to ensure adequate N uptake and dry matter production as the fertilizer diffuses and the root system extends. In subsequent growth stages, mineral N availability would fully satisfy the growth demands of rice. The total N uptake was higher in the middle season than in the early and late seasons because of the differences in growth and variety characteristics of Yongxian15 and Yongyou1540 (Fig. 5). In all three seasons, deep placement of N fertilizer (CRUM and UM) significantly increased N uptake across all growth stages and increased the NRE, compared with those in UB. This implied that N losses were lower in CRUM and UM than in UB. The deep application of N fertilizer (UM and CRUM) increased N uptake by: (1) reducing N 2 O emission 47 31,42,43 ; and (4) synchronizing N availability with plant N demands. The N uptake from the panicle initiation stage to the maturity stage was significantly higher in the mechanically applied N fertilizer treatments (irrespective combinations of N fertilizer applied) than in the manual surface broadcasting treatment.
The N nutrition of rice plants depends on the physical, chemical, and biochemical transformations of N, and on the diffusion dynamics and the forms of N derived from fertilizer and other N sources in soil. The extension and development of the rice root system also affect the N nutrition of rice plants. The NRE and NAE were higher in CRUM and UM than in UB in the early, middle, and late seasons, consistent with previous findings 34,37,38,40,48 . In this study, the deep placement of N fertilizer greatly increased NAE and NRE, regardless of the type of N fertilizer applied ( Table 2). In previous studies, the NUE was found to be affected by the N type and the type of pellets used 20,37,41,44 . The higher NRE and NAE under UM were attributed to greater N uptake that was mainly allocated to rice grains. The NRE and NAE were higher in CRUM than in UM in all three seasons, consistent with previously reported results 44 . The N uptake of rice in UM and CRUM were enhanced because of the higher + NH 4 -N and − NO 3 -N concentrations in soil (especially closer to the roots) during the whole rice growth period. Our results also showed that greater pre-heading dry matter accumulation and N uptake were essential to enhance grain yield and NUE. This is because the increased N uptake before the heading stage in UM and CRUM improved N translocation and dry matter production during the grain filling stage, as reported in another study.
Compared with manual surface broadcasting, the mechanized side deep fertilization method had much lower fertilization costs, regardless of the type of fertilizer used (Table 3). Thus, the fertilizer application method was the major cost factor among the inputs. Great savings in labour costs could be made by simultaneously transplanting and fertilizing rice because of the high operational efficiency of this method 33,37,49 . It was reported that in various locations and crop production seasons, the lack of available labour for manual surface broadcasting of fertilizer could result in a crisis 5,7,50 . The rapid economic development in China has meant that the rural labour force has become much smaller as workers have moved to developed coastal areas or cities. Therefore, the rural labour force (2019) 9:5653 | https://doi.org/10.1038/s41598-019-42039-7 www.nature.com/scientificreports www.nature.com/scientificreports/ is becoming scarcer and more expensive. To address this issue, the mechanical side deep fertilization of mechanically transplanted rice can be up-scaled in rice production systems.
In our study, the grain yield in the early, middle and late seasons was highest in CRUM, followed by UM and UB. The BCR and net income were highest in CRUM, followed by UM, and lowest in UB. The combination of higher gross return and lower fertilization cost in CRUM resulted in higher BCR and net income, consistent with previous results 37,42 . In those studies, grain yield and net economic return were increased by deep fertilization of either urea briquettes or N-P-K briquettes compared with broadcasting prilled (pelletized) urea, while deep fertilization of either urea briquettes or N-P-K briquettes reduced N fertilizer use and increased rice production. Other studies have shown that deep placement of fertilizer can increase nutrient-use efficiency 51,52 , thereby reducing the fertilizer requirements without yield penalties. In those studies, basal N fertilizer (40-60% N) was applied by a deep mechanical method or by manual broadcasting, and remaining fertilizer was applied by manual broadcasting. The urea briquette applicator used at the first top and second top dressings resulted in a 42.8% time saving over hand application 42 , which significantly reduced operation times and management costs. Considering that the CRUM method in this study required only two fertilizer applications, increased N availability in paddy fields, and increased grain yields and NUE (Tables 1 and 2), local farmers should be encouraged to adopt this technology. Further research is required to determine whether there are additional benefits of supplying all primary nutrients and balancing fertilization, and to determine whether the mechanized single-dose deep application of a controlled release compound fertilizer (N-P-K) would result in lower labour costs and better nutrient-use efficiency.

Conclusion
Compared with manual surface broadcasting of N fertilizer, mechanized side deep N fertilization significantly increased the NUE and grain yield of machine-transplanted rice. After the heading stage, the Pn, total biomass, total N uptake of rice plants were higher in the treatments with mechanically applied fertilizer than in the treatments with manually broadcasted urea. Mechanized side deep fertilization of U and CRU (1:1) also resulted in the highest BCR because of higher total returns and the lower fertilization costs. These results show that mechanized deep side fertilization is an effective technology not only to reduce labour costs and nutrient losses, but also to increase economic profitability. Continuing advancements in such fertilization and production technologies for controlled-release compound fertilizer have the potential to greatly reduce the amount of chemical fertilizer and labour costs, and improve grain yields and resource use efficiency for rice growers worldwide.

Materials and Methods
Side deep fertilizer applicator. The machine for the wide-narrow row transplanting of rice accompanied by side deep fertilization was developed by the Jinhe Agricultural Science and Technology Co., Ltd. (Zhejiang, China) (Fig. 1a).
Experimental site and weather conditions. The field experiments were conducted in three growing seasons: the early season from March to July; the middle season from April to October; and the late season from July to November. All field experiments were conducted in 2017 at the Experimental Research Farm, Langxin Food Professional Cooperatives, Langya Town, Jinhua City, Zhejiang Province, China (119.47°E, 29.02°N, elevation 81 m above sea level). In general, this region has a monsoon and sub-tropical climate (for details, see Table 4). The physicochemical properties of the soil before starting the experiments are shown in Table 5. Soil pH was measured in 1:2.5 (v/v) soil to water ratio using a pH meter. Soil total N, total P, and total K contents were measured by the Kjeldahl method, colorimetric analysis, and flame photometry, respectively. Soil organic matter was determined by the wet combustion method. Experimental material. We used two types of widely grown rice cultivars, Yongxian15 (Indica conventional rice) and Yongyou1540 (Indica/Japonica hybrid rice), which were supplied by the Ningbo Academy of Agricultural Sciences, Zhejiang, China and Ningbo Seed Co., Ltd., China, respectively. Yongxian15 (early season rice) had a growth period of 107.9 days. Yongyou1540 had growth periods of 151.5 and 123.9 days as middle-season and late-season rice, respectively. The two types of N fertilizers were common urea (U, total N www.nature.com/scientificreports www.nature.com/scientificreports/ content TN = 46.7%) and controlled release urea (CRU, total N content TN = 41.6%, with a slow release period of 60 days), both of which were manufactured by Sinofert Holdings Limited, China.
Experimental design and treatments. Two fertilization methods were used: manual surface broadcast (B), and 5.5 ± 0.5 cm-depth and horizontal width mechanized side deep placement (M). Four N-fertilizer treatments were arranged in a randomized complete block design with three replications in each season. The experimental plots were 120 m 2 in the early rice season and 100 m 2 in the middle and late rice seasons. The N-fertilizer treatments were as follows: Early rice season. UB: 144 kg N ha −1 as U was broadcasted manually onto the soil surface as seedling fertilizer, and 36 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. UM: 144 kg N ha −1 as U was mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 36 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. CRUM: 72 kg N ha −1 as U and 72 kg N ha −1 as CRU were mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 36 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage.
Middle rice season. UB: 192 kg N ha −1 as U was broadcasted manually onto the soil surface as seedling fertilizer, and 48 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. UM: 192 kg N ha −1 as U was mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 48 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. CRUM:96 kg N ha −1 as U and 96 kg N ha −1 as CRU were mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 48 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage.
Late rice season. UB: 168 kg N ha −1 as U was broadcasted manually onto the soil surface as seedling fertilizer, and 42 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. UM: 168 kg N ha −1 as U was mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 42 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage. CRUM: 84 kg N ha −1 as U and 84 kg N ha −1 as CRU were mechanically fertilized at 5.5 ± 0.5 cm soil depth beside seedlings when transplanting, and 42 kg N ha −1 as U was broadcasted manually onto the soil surface at the panicle initiation stage.
Plots with no added N fertilizer were established as the blank control (N0) in each season. Each treatment received 90 kg P 2 O 5 ha −1 as superphosphate and 120 kg K 2 O ha −1 as potassium chloride. All phosphate fertilizer and 50% of potash fertilizer were broadcasted onto the soil surface 1 day before transplanting, and 50% of potash fertilizer was applied at the panicle initiation stage in all treatments. In all rice seasons, the strategy for water management was the sequence of flooding, midseason drainage, re-flooding, moist intermittent irrigation, and drainage. Weeds, insects, and diseases were intensively controlled by chemicals. Other field practices, such as field preparation, tillage, and puddling, were carried out manually according to the local farming practices.
Total above-ground biomass and leaf area index (LAI). Six hills of plants were sampled from each plot according to average tiller number at the panicle initiation, heading and maturity stages. The adhered soil was thoroughly washed from the plants, and then the panicles, leaves, and stems with leaf sheaths were cut away from the plants after the heading stage. A LI-3100C Area Meter (LI-COR, Inc., USA) was used to measure the leaf area of each green leaf, then the leaf area per square meter was calculated as LAI. To record the total above-ground biomass, the sampled plants were dried at 105 °C for 30 min to halt biological activity, and then dried at 80 °C to constant weight (DHG-9625A, Shanghai Yiheng Scientific Instruments Co., Ltd., Shanghai, China).
Net photosynthetic rate (Pn). At the heading stage, LI-6400XT Portable Photosynthesis System (LI-COR, Inc., Lincoln, NE USA) was used to measure Pn of the flag leaves in the early, middle, and late seasons in 2017. The photosynthetically active radiation was controlled at 1200 μmol m −2 s −1 provided by a 6400-2B LED light  Table 5. Physicochemical properties of soil in early season, middle season, and late season of 2017.
www.nature.com/scientificreports www.nature.com/scientificreports/ source. Six representative flag leaves from each plot were measured and recorded, and the mean values were calculated for Pn.
Yield and its components. At the maturity stage, six hills of plants were collected from each plot to measure yield components. The straw and panicles were cut away from the plants. To separate filled grains, each grain was separated from the rachides of the spike (through manual threshing) and immersed in distilled water. Those that floated were considered to be unfilled grains. To measure the gross number of spikelets, we counted spikelets in three representative subsamples of 30 g. The average weight of half-filled spikelets was determined. The grain number per panicle, the percentage of grain filling, and the 1000-grain-weight were measured according to Pan S. et al. 37 . Rice plants from a 60.0 m 2 area were harvested in plots and the grain yield was calculated based on a standardized moisture content of 14%. Plant N content, N uptake, and N use efficiency. At the panicle initiation, heading and maturity stages, we collected panicles, leaves, stems with leaf sheath samples from each plot to analyse their total N contents. The plant samples (0.20 g) were digested for 2 h in H 2 SO 4 -H 2 O 2 solution at 420 °C and analysed by the micro-Kjeldahl method (Kjeltec TM 8400, FOSS, Denmark).
N uptake was calculated using the formula TDM × NC, where TDM represents the total dry matter of panicles, leaves, and stems with leaf sheaths, and NC represents the N concentration in panicles, leaves, and stems with leaf sheaths.
Aspects of N use efficiency, such as N dry matter production efficiency (NDMPE), N grain production efficiency (NGPE), N recovery efficiency (NRE), N agronomic use efficiency (NAE), and N physiological efficiency (NPE) were calculated as the following formulas: where TBup and TNup denote the total above-ground biomass and total N uptake above-ground, respectively; GY and GY0 represent grain yields in N-fertilized plots and N0 plots, respectively; Nup and N0up denote total N uptake above-ground in N-fertilized plots and N0 plots, respectively; FN denote the total N application rate in N-fertilized plots. Economic profitability. Economic profitability was calculated from the rice production costs and the net income obtained from the rice sale price. The rice production costs consisted of fertilizer cost, fertilization cost, other input. Fertilizer cost included the expenditures on N fertilizer, phosphate fertilizer, potash fertilizer. Fertilization cost included the expenditures on labour and fuel. Other input included the expenditures on seed, rotary tillage, planting and management (labour and fuel on raising rice seedlings, transplanting, spraying pesticides, and water management), farm pesticides, harvest, transportation, and drying. All of these costs were the local average price in the rice growing seasons. Net income was estimated by subtracting various costs and expenses from the calculated gross return. The ratio of gross return to total production cost was calculated as the benefit cost ratio. Data analysis. Data are presented as means ± SE of three plots (three replicates). The dataset for early, middle, and late season were analysed using SAS ® 9.1. Pairwise means comparison among treatments was conducted with least significant difference (LSD) tests at the 0.05 level of probability.

Data Availability
The data used or analysed during the current study are available from the corresponding author on reasonable request.