lmproving grain yield,nitrogen use efficiency and radiation use efficiency by dense planting,with delayed and reduced nitrogen application,in double cropping rice in South China

FU You-qiang,ZHONG Xu-hua,ZENG Jia-huan,LIANG Kai-ming,PAN Jun-feng,XIN Ying-feng,LlU Yan-zhuo,HU Xiang-yu,PENG Bi-lin,CHEN Rong-bing,HU Rui,HUANG Nong-rong

1 Rice Research Institute,Guangdong Academy of Agricultural Sciences/Guangdong Key Laboratory of New Technology for Rice Breeding,Guangzhou 510640,P.R.China

2 Guiping Agricultural Technology Extension Station,Guiping 537200,P.R.China

Abstract Improving both grain yield and resource use efficiencies simultaneously is a major challenge in rice production. However,few studies have focused on integrating dense planting with delayed and reduced nitrogen application to enhance grain yield,nitrogen use efficiency (NUE) and radiation use efficiency (RUE) in rice (Oryza sativa L.) in the double rice cropping system in South China. A high-yielding indica hybrid rice cultivar (Yliangyou 143) was grown in field experiments in Guangxi,South China,with three cultivation managements:farmers’ practice (FP),dense planting with equal N input and delayed N application (DPEN) and dense planting with reduced N input and delayed N application (DPRN). The grain yields of DPRN reached 10.6 and 9.78 t ha–1 in the early and late cropping seasons,respectively,which were significantly higher than the corresponding yields of FP by 23.9–29.9%. The grain yields in DPEN and DPRN were comparable. NUE in DPRN reached 65.2–72.9 kg kg–1,which was 61.2–74.1% higher than that in FP and 24.6–30.2% higher than that in DPEN. RUE in DPRN achieved 1.60–1.80 g MJ–1,which was 28.6–37.9% higher than that in FP. The productive tiller percentage in DPRN was 7.9–36.2% higher than that in DPEN. Increases in crop growth rate,leaf area duration,N uptake from panicle initiation to heading and enhancement of the apparent transformation ratio of dry weight from stems and leaf sheaths to panicles all contributed to higher grain yield and higher resource use efficiencies in DPRN. Correlation analysis revealed that the agronomic and physiological traits mentioned above were significantly and positively correlated with grain yield.Comparison trials carried out in Guangdong in 2018 and 2019 also showed that DPRN performed better than DPEN. We conclude that DPRN is a feasible approach for simultaneously increasing grain yield,NUE and RUE in the double rice cropping system in South China.

Keywords:grain yield,resource use efficiencies,indica rice,planting density,nitrogen application strategy

1.lntroduction

Nitrogen (N) is one of the main factors which has contributed greatly to the improvement of rice grain yield in the past half century (Chenet al.2014; Juet al.2015). The planting density is another important factor,which directly affects the panicle number,and thereby influences the rice grain yield (Huanget al.2011; Liet al.2011). At present,in order to save more labors and ensure higher grain yield,most rice farmers are used to planting rice with a sparse density and high N input (Penget al.2006; Chenet al.2015). The reason is that grain yield losses due to the sparse planting density can be compensated by the earlier tillering and fast growth under the heavy N input (Linet al.2009; Huanget al.2013). However,the average N fertilizer input has reportedly reached 180 kg ha–1for irrigated rice in China,which is approximately 75% higher than the world average(FAO 2012),and 50–90% of the N fertilizer is applied as basal dressing and early topdressing within the first 10 days after transplanting (Penget al.2006; Zhanget al.2011).Moreover,the sparse planting density reduces the rice’s ability to compete with weeds at the early growth stages owing to the increased space that is available for the growth of the weeds (Parry and Hawkesford 2010). The large amounts of N fertilizer applied at the early growth stages and the sparse planting density not only result in a mismatch between N application and crop demand,higher N loss,water eutrophication and lower grain yield,but also stimulate weed overgrowth (Zhuet al.2017). Therefore,finding the optimized planting density and N application strategies is very important for improving rice grain yield and maintaining environmental sustainability (Yousafet al.2016).

Rice (Oryza sativaL.) is one of the most important staple food crops in the world,providing food for more than 50%of the world’s population,and it plays an important role in maintaining food security (Yuan 2014). By the year 2035,the annual grain production will increase by an estimated 116 million tons to feed the ever-increasing population(Yamanoet al.2016). In the second half of the last century,world rice production was dramatically increased,primarily as the result of improvements in breeding and cultivation technologies (Penget al.1999; Zhanget al.2011).However,in the past decade,the increase of rice grain yield has slowed down despite of the increasing fertilizer input in China and other countries. This has resulted mainly from the excessive and inappropriate timing of fertilizer application (Katsuraet al.2007; Normile 2008). Therefore,increasing crop production to maintain food security and enhancing resource use efficiencies without increasing the environmental impacts are major challenges (Wuet al.2017; Jiaoet al.2018).

Many optimized N management strategies,such as sitespecific N management (Penget al.2010),deep placement of fertilizer (Liet al.2014),precise quantitative cultivation techniques (Linget al.2005),“three controls” nutrient management technology (Zhonget al.2007),and others,have been applied to increase rice yield and nitrogen use efficiency (NUE). Most of these technologies can obtain higher grain yields and NUE than the standard farmers’practices (FP) (Penget al.2006; Miaoet al.2011). In addition,previous studies have indicated that increasing planting density can increase panicle number,which can alleviate the negative effect on yield of reducing the N rate (Tianet al.2017; Houet al.2019). Tianet al.(2017)proposed that adjusting the planting density could be an efficient method for reducing the amount of N fertilizer and improving the NUE. Our previous studies have revealed that “three controls” nutrient management technology,with both reduced N input and delayed topdressing time under medium planting density conditions,could reduce the occurrence of unproductive tillers and increase both grain yield and NUE in the double rice cropping system in South China (Zhonget al.2010; Lianget al.2019; Panet al.2019).Therefore,planting density and N application strategies are the two most important crop management practices that significantly influence the grain yield and NUE of irrigated rice (Huanget al.2013; Ahmedet al.2016).

Many studies have reported the effect of simultaneous changes in the planting density and N application strategies on grain yield and NUE,especially in the single-season rice cropping system in China (Chenet al.2019; Houet al.2019; Xieet al.2019). However,in the doubleindicarice cropping system of South China,it remains largely unknown whether optimizing both planting density and N application strategies can improve rice grain yield,NUE and radiation use efficiency (RUE) simultaneously. The objectives of the present study were:(i) to compare the differences in the desirable agronomic traits at different growth phases among various cultivation management strategies,(ii) to find a way to narrow the gaps of grain yield,NUE and RUE between the optimized cultivation managements and FP and (iii) to understand the mechanism underlying the simultaneous improvements in grain yield and resource use efficiencies achieved by optimizing both planting density and N application strategies in the double rice cropping system of South China.

2.Materials and methods

2.1.Site description

Field experiments were conducted in Shilong Township(23°19´N,109°53´E,44 m a.s.l.),Guiping County,Guangxi Zhuang Autonomous Region of South China in 2019,and comparison trials were carried out in Xinxu Township(24°3´N,115°52´E,115 m a.s.l.),Xingning County,Guangdong Province of South China during 2018 and 2019.

Soil samples were collected from the 0 to 20 cm soil depth for analysis of soil properties prior to the experiments.The physico-chemical properties of the soil in the field experiments (Guiping site) were a clay loam texture with pH 5.67,organic matter 25.38 g kg–1,total N 1.43 g kg–1,available P 54.10 mg kg–1,and K 94.86 mg kg–1; and in the comparison trials (Xingning site),there were a clay loam texture with pH 5.33,organic matter 28.82 g kg–1,total N 1.47 g kg–1,available P 2.48 mg kg–1,and K 56.68 mg kg–1.

The data of daily mean temperature,rainfall and solar radiation during the rice growing period were collected at Guiping National Meteorological Observatory (station number:59254,23°39´N,110°06´E,44 m a.s.l.) near the field experiment site in 2019,and Wuhua National Meteorological Observatory (station number:59303,23°55´N,114°45´E,135 m a.s.l.) near the comparison trial site during 2018 and 2019. These data are shown in Fig.1.With regard to the Guiping site,the average temperatures,seasonal rainfall and total solar radiation from transplanting to harvest were 26.3 and 25.6°C,732.2 and 474.5 mm,and 1 321.8 and 1 732.4 MJ m–2in the early and late seasons in 2019,respectively (Fig.1-A and B). With regard to the Xingning site,the average temperatures were 25.8 and 24.7°C in 2018,and 25.2 and 25.4°C in 2019; seasonal rainfall was 472.8 and 466.6 mm in 2018,and 992.6 and 333.5 mm in 2019; and total solar radiation was 1 990.5 and 1 560.1 MJ m–2in 2018,and 1 785.1 and 2 062.5 MJ m–2in 2019,in the early and late seasons,respectively (Fig.1-C–F).

2.2.Experimental design and treatments

Experimental designPrior to the field experiment,a preliminary comparison trial was conducted in the early and late seasons to explore the effect of optimizing both planting density and N application strategies on grain yield,NUE and RUE at the Xingning site in 2018. To further understand the mechanism underlying the simultaneous improvement of these parameters by optimizing both planting density and N application strategies,a field experiment and a comparison trial were subsequently carried out at Guiping and Xingning,respectively,in 2019.

The plots were arranged in a randomized complete block design with three replications. Plot size was 105 m2at the Guiping site and 600 m2at the Xingning site,and each plot was separated by a ridge of 0.5 m width with plastic film inserted into the soil to a depth of 0.3 m to minimize leakage and nutrient loss.

At the Guiping site,there were three treatments:(1) FP:farmer’s practice,with planting density 18.5 hills m–2,hill spacing 30 cm×18 cm,and N application rate 195 kg N ha–1; (2) DPEN:dense planting (28 hills m–2,30 cm×12 cm,51% greater than FP) with equal nitrogen rate (195 kg N ha–1,the same as FP) and delayed N application; and(3) DPRN:dense planting (28 hills m–2,30 cm×12 cm) with reduced nitrogen rate (150 kg N ha–1,23% lower than FP)and delayed N application.

At the Xingning site,there were two treatments of DPEN and DPRN as described above. Rice seedlings were transplanted with 3–5 seedlings per hill by a mechanical rice transplanter in the 2 years (Table 1).

The DPEN and DPRN treatments were designed based on the principle of “three controls” nutrient management technology (Zhonget al.2007,2010). The comparison trials were treated the same as the field experiments with respect to nutrient management.

Experimental treatmentsIn the FP in South China,the total N rate of 195 kg ha–1consisted of 51,34,69,and 41 kg N ha–1applied at the basal,recovering (3–5 days after transplanting),early tillering (7–10 days after transplanting),and late tillering (20 days after transplanting) stages,respectively. Rice seedlings were transplanted at 18.5 hills m–2. Potassium (105 kg K2O ha–1,potassium chloride) was split equally and applied at both the basal and early tillering stages.

In the treatment of DPEN,the total N rate of 195 kg ha–1(the same as in FP) consisted of 76,49,35,and 35 kg N ha–1applied at the basal,mid tillering (15–18 days after transplanting),panicle initiation,and booting stages,respectively. The planting density was 28 hills m–2.Potassium (105 kg K2O ha–1,potassium chloride) was split equally and applied at both mid tillering and panicle initiation.

In the treatment of DPRN,the total N rate of 150 kg ha–1was applied at the basal,mid tillering and panicle initiation stages at rates of 69,34 and 47 kg ha–1,respectively,in both years. The planting density was 28 hills m–2. K2O (105 kg ha–1,potassium chloride) was split equally and applied at both mid-tillering and panicle initiation.

Urea (46% N) was used as the N source in these experiments. P2O5(30 kg ha–1as calcium superphosphate)was applied as basal fertilizer in each plot before transplanting.

Plant material and crop managementAnindicahybrid rice cultivar,Yliangyou 143 (Oryza sativaL.),with high yield and disease resistance,which is widely grown in South China,was used in the paddy field experiments for both years. At the Guiping experiment site in 2019,pre-germinated seeds were sown on 28 February in the early season and 14 July in the late season,and 30-d-old seedlings were transplanted on 1 April in the early season and 20-d-old seedlings were transplanted on 3 August in the late season. At the Xingning site in 2019,pre-germinated seeds were sown on 10 February in the early season and 12 July in the late season,and 47-d-old seedlings were transplanted on 29 March in the early season and 25-d-old seedlings were transplanted on 6 August in the late season. At the Xingning site in 2018,the pre-germinated seeds were sown on 21 February in the early season and 17 July in the late season,and 41-d-old seedlings were transplanted on 3 April in the early season and 24-d-old seedlings were transplanted on 10 August in the late season. A water depth of 2–5 cm during the first 10 d after transplanting was maintained to facilitate tillering,and then at around 25 days after transplanting,the mid-season drainage was conducteduntil panicle initiation,and subsequent shallow wetting irrigation was carried out until 7 d before harvest. Diseases and insects were intensively controlled by chemicals (as needed),and weeds were removed by hand to avoid yield losses during the entire growing season; no obvious water,weed,disease,or pest stresses were observed at either the Guiping or Xingning sites during the experiments in either year.

2.3.Plant sampling and measurements

Measurement of agronomic traits before harvestTwelve hills on a diagonal from each plot were sampled to record panicle number (main stems plus tillers,a tiller with at least one leaf was counted as a stem) at the transplanting,midtillering and panicle initiation stages. The sampled plants were separated into leaves and stems. At the heading stage,the sampled plants were partitioned into leaves,stems (culm+sheath) and panicles. The green leaf area was measured using a LI-3100C leaf area meter (LI-COR,Inc.Lincoln,NE,USA) and leaf area index (LAI) was calculated as the green leaf area per unit land area. The dry weight of each part was determined after oven-drying at 75°C to constant weight. Total dry matter weight was the sum of the weights for each part in a given stage. Leaf area duration(LAD) and crop growth rate (CGR) were calculated using the following formulas:LAD (m2m–2d)=1/2(L1+L2)(T2–T1); and CGR (g m–2d–1)=(W2–W1)/(T2–T1); where L1and L2are the first and second measurements of LAI (m2m–2),respectively,W1and W2are the first and second measurements of shoot biomass (g m–2),respectively,and T1and T2represent dates of the first and second measurements (d),respectively(Zhanget al.2018).

Determination of agronomic traits and grain yield after harvestAt maturity,12 hills on a diagonal from the center of each plot were sampled to determine the yield components. The sampled plants were separated into straws,filled spikelets and unfilled spikelets (including rachis). Grain yield was determined from 4.5 m2(125 hills)harvest areas for DPEN and DPRN,and from 6.75 m2(125 hills) for FP due to difference of planting density (except border plants) in each plot,and yields were then adjusted to the standard moisture content of 0.14 g H2O g–1fresh weight. Panicle number was recorded from 12 hills. The panicles were hand-threshed and the filled spikelets were separated from unfilled spikelets using a seed winnowing machine (FJ-I,Hangzhou,China). The numbers of filled and unfilled grains were calculated using an automatic seed counter (TPZJ-A,Hangzhou,China). Dry weights of straw,filled and unfilled spikelets were measured after oven-drying at 75°C to constant weight. Spikelets per panicle and seed set rate (100×Filled spikelet number/Total spikelet number)were calculated. Harvest index (HI) was calculated as the ratio of filled spikelets weight to total biomass. Spikelets to leaf-area ratio at heading stage (HD) was calculated using the formula:Spikelets to leaf-area ratio=Spikelets m–2at maturity stage (MA)/LAI at HD. Sink size was calculated as the product of spikelets m–2and grain weight. The apparent transformation ratio (%) was calculated as:(Dry weight (leaf and stem) at heading stage–Dry matter weight (straw) at maturity)/Panicle dry weight at maturity×100 (Yuet al.2015).

Determination of tissue N concentrations and NUEThe dried samples of each part (leaf and stem at transplanting,mid tillering and panicle initiation; leaf,stem and panicle at heading; and straw,and filled and unfilled spikelets at maturity) were ground separately and respectively to a fine powder to determine the tissue N concentration using micro-Kjeldahl digestion,distillation and titration (Yoshida and Parao 1976). Total N uptake at each stage was calculated as the summation of the individual N uptakes of each part.The partial factor productivity of applied N fertilizer (PFPN,kg kg–1) was calculated as:grain yield (kg)/N rate (kg). In the present study,PFPNis used as NUE.

Measurement of canopy light interception and RUEThe proportion of canopy light interception was measured between 11:00 and 14:00 h at the mid tillering,panicle initiation,heading,and maturity stages,using an AccuPAR LP-80 (Decagon Devices,Inc.Pullman,WA,USA). In each plot,a 1.0-m line sensor was placed slightly above the water surface in the middle of two rows,and a dot sensor was placed above the crop canopy. Five readings of incoming light intensity and light intensity inside the canopy were recorded simultaneously.Canopy light interception(%) was calculated as:100×(1–Average light intensity inside canopy/Incoming light intensity) (Zhanget al.2009).Incident radiation was collected using an automatic solar radiation sensor and data logger (H21-USB+S-LIB-M003,Onset HOBO,USA) placed next to the experimental plots. Intercepted radiation during each growth period was calculated as:1/2×(Canopy light interception at the beginning of the growth period+Canopy light interception at the end of the growth period)×Accumulated incoming radiation during the growth period. RUE was calculated as the ratio of aboveground total dry weight to accumulated intercepted radiation over the period of each growth stage(Trápaniet al.1992).

2.4.Statistical analysis

Analysis of variance (ANOVA) and correlation analysis(Pearson) were performed using Statistix 8 (Analytical Sorftware,Tallahassee,FL,USA). Data from each sampling date were analyzed separately. Means were tested by the least significant difference (LSD) test at the 5% level.The figures were generated using Origin Pro 8.0 Software(OriginLab,Northampton,MA,USA).

3.Results

3.1.Grain yield and its components

In the field experiments,significant differences in grain yield,panicle number m–2,seed set rate,and sink size were determined among the treatments by ANOVA. No significant differences were observed among treatments with regard to spikelets per panicle,1 000-grain weight or HI. There were no significant season-treatment interactive effects for any of the traits in 2019 (Table 2). Grain yield in FP was 8.16 t ha–1in the early season and 7.89 t ha–1in the late season. Compared to FP,grain yields were higher by 29.66 and 23.57% for DPEN,and by 29.90 and 23.95%for DPRN in the early and late seasons,respectively. No significant difference in grain yield was observed between DPEN and DPRN.

With regard to yield components,the panicle numbers under DPEN and DPRN were increased markedly regardless of early and late season timing,by 23.4–26.5%and 26.1–37.4%,respectively,compared with that under FP.In addition,DPRN and DPEN significantly improved sink size and HI compared with FP,except for the HI in DPEN in the early season. The DPEN treatment significantly increased spikelets per panicle but decreased seed set rate compared with DPRN in the early season. In contrast,DPRN increased spikelets per panicle but significantly reduced 1 000-grain weight compared with DPEN in the late season,whereas there were no significant differences in seed set rate,sink size or HI between DPEN and DPRN.

3.2.Number of tillers and the percentage of productive tillers

The number of tillers varied with various treatments and measurement stages (Fig.2). Compared to FP,DPEN increased the number of tillers by 14.4,17.6,72.4,and 23.4% in the early season and by 20.2,11.5,42.6,and 26.5% in the late season,at the mid tiller,panicle initiation,heading and maturity stages,respectively. DPRN increased the number of tillers by 9.2,11.4,56.6,and 26.1% in the early season and by–5.1,–3.2,32.9,and 37.4% in the late season,respectively,at the corresponding stages mentioned above. The percentages of productive tillers for DPEN and DPRN were increased by 4.6 and 12.9% in the early season and by 6.1 and 44.5% in the late season,respectively,in comparison with FP. The productive tiller percentage under DPRN was 7.9 and 36.2% higher than that of DPEN,in the early and late seasons,respectively,and the differences were significant.

3.3.Biomass and CGR

The dry matter accumulation of each rice growth phase gradually increased from transplanting (TP) to mid tillering(MT) and from MT to panicle initiation (PI),reaching a peak from the PI to heading stage (HD),and declining thereafter in both the early and late seasons (Fig.3-A and B). Although no significant differences in dry matter accumulation among the three treatments from MT to PI or from HD to maturity(MA) were observed in either the early and late seasons,it was significantly increased by 13.8 and 63.2% in the early season and by 32.1 and 26.1% in the late season under the DPEN treatment compared to FP,and by 21.8 and 46.7% in the early season and by 26.8 and 35.8% in the late season under the DPRN treatment in comparisonwith FP,respectively,from TP to MT and from PI to HD.Consequently,DPEN and DPRN significantly increased the dry matter weight of the whole growth period (WGP) by 34.4 and 32.5% in the early season and by 26.4 and 46.4% in the late season compared with FP,respectively. A similar tendency was found for CGR (Fig.3-C and D).

Table 2 Grain yield and its components under different treatments in the early and late seasons in 2019 in Guiping,Guangxi Zhuang Autonomous Region of China

Fig.2 Number of tillers (A and B) and the percentage of productive tillers (C and D) under various treatments at Guiping site,Guangxi,China during 2019 early (A and C) and late (B and D) seasons. MT,mid tillering; PI,panicle initiation; MA,maturity; TP,transplanting. FP,farmer’s practice; DPEN,dense planting with equal N input and delayed N application; DPRN,dense planting with reduced N input and delayed N application. Percentage of productive tillers was calculated as the ratio of tillers at maturity to the maximum number of tillers during rice growth period. Bars are mean±SE (n=3). Different letters indicate statistical significance at the P=0.05 level within the same stage and the same season.

3.4.The performance indicated by other agronomic traits

Significant differences were found in the traits of LAD,dry weight of green leaves,spikelets to leaf-area ratio,and apparent transformation ratio among the treatments (Fig.4).Compared to FP,between PI and HD under DPEN and DPRN conditions,LAD was increased significantly by 39.0 and 36.0% in the early season and by 23.6 and 19.3% in the late season; the dry weight of green leaves at HD under DPEN and DPRN was increased by 73.1 and 46.9% in the early season and by 39 and 30.3% in the late season; while the apparent transformation ratio under DPEN and DPRN was increased by 47.3 and 49.8% in the early season,and by 49.5 and 52.8% in the late season,respectively.Compared to FP,the spikelets to leaf-area ratio under DPEN and DPRN was decreased by 21.3 and 11.5% in the early season,respectively. While the spikelets to leaf-area ratio under DPRN was 45.4% higher than that under FP,no difference was observed between those ratios under DPEN and FP in the late season.

3.5.Nitrogen uptake and leaf N concentration

The results of N uptake at each growth phase and leaf N concentrations at HD in various treatments are shown in Fig.5. The N uptake was increased in both growth phases from TP to MT and from MT to PI,and peaked from PI to HD,then declined across both seasons. Compared to that under FP,the N uptake from TP to MT and from PI to HD was increased by 13.0% and 4.05-fold in the early season and by 8.4% and 2.6-fold in the late season under DPEN,and by 19.8% and 2.0-fold in the early season and by 6.2% and 2.4-fold in the late season under DPRN,respectively. The N uptake from HD to MA under DPRN was increased by 1.9-fold in late season compared to that under FP. Consequently,N uptake of the whole growth period under DPEN and DPRN was increased significantly compared with FP,by 66.1 and 39.0% in the early season and by 41.8 and 54.5% in the late season,respectively. With regard to leaf N concentration,DPEN and DPRN increased the leaf N concentration at HD by 44.2 and 20.8% in the early season,and by 29.0 and 17.6%in the late season,respectively,compared with FP.

Fig.3 Dry matter accumulation (A and B) and crop growth rate (CGR) (C and D) for each rice growth phase under various treatments at Guiping site,Guangxi,China during 2019 early (A and C) and late (B and D) seasons. FP,farmer’s practice; DPEN,dense planting with equal N input and delayed N application; DPRN,dense planting with reduced N input and delayed N application.TP,transplanting; MT,mid tillering; PI,panicle initiation; HD,heading stage; MA,maturity; WGP,the whole growth period of rice.Bars are mean±SE (n=3). Different letters indicate statistical significance at the P=0.05 level within the same growth phase and the same season.

3.6.Nitrogen and radiation use efficiencies

The results of the RUE and nitrogen partial factor productivity(PFPN, as expressed NUE) of various treatments are shown in Fig.6. No significant difference was observed in RUE among the three treatments from TP to MT,from MT to PI or from HD to MA in the early season,or from MT to PI in the late season. In the early season of 2019,the RUE of DPEN and DPRN from PI to HD were 63.1 and 53.6% greater than that of FP,respectively. In the late season of 2019,the RUE was increased by 22.6,22.6 and 35.6% under DPEN,and by 25.8,35.9 and 123.8% under DPRN relative to that under FP,from TP to MT,from PI to HD and from HD to MA,respectively. Accordingly,DPEN and DPRN increased the RUE of the whole growth period compared with FP by 35.0 and 28.6% in the early season and by 21.6 and 37.9% in the late season,respectively. Interestingly,the RUE was increased significantly by 10.8,65.0 and 13.5% in DPRN relative to DPEN in the late season,from PI to HD,from HD to MA and across whole growth period,respectively.

With regard to NUE (Fig.6),the NUE of DPEN and DPRN increased significantly by 33.7 and 74.1% in the early season and by 29.4 and 61.2%,respectively,in the late season in comparison with FP. Furthermore,the NUE of DPRN was significantly higher than that of DPEN,regardless of early or late season timing.

3.7.The relationships between grain yield and agronomic and physiological traits

Fig.4 Leaf area duration (LAD) from panicle intiation stage (PI) to heading stage (HD) (A and B),dry weight of green leaves at HD(C and D),spikelets to leaf-area ratio (E and F),and apparent transformation ratio (G and H) under various treatments at Guiping site,Guangxi,China during 2019 early (A,C,E,and G) and late (B,D,F,and H) seasons. FP,farmer’s practice; DPEN,dense planting with equal N input and delayed N application; DPRN,dense planting with reduced N input and delayed N application. Bars are mean±SE (n=3). Different letters indicate statistical significance at the P=0.05 level within various treatments.

As shown in Fig.7,correlation analysis revealed that grain yield had positive linear relationships with panicle number and total dry weight (P<0.01),while there was no correlation between grain yield and either spikelets per panicle (P>0.05),seed set rate (P>0.05),1 000-grain weight (P>0.05) or HI(P>0.05). In addition,grain yield was positively correlated with LAD from PI to HD (P<0.01),with CGR between PI and HD (P<0.01),with apparent transformation ratio(P<0.01) and leaf nitrogen concentration at HD (P<0.01),and with NUE (P<0.01) and RUE for the whole growth period (P<0.01).

3.8.Performance of DPEN and DPRN in South China during 2018 and 2019

A comparison trial with DPEN and DPRN treatments was conducted to investigate the effect of optimizing both planting density and N application on grain yield,NUE and RUE in double cropping rice in Xingning County,Guangdong Province of South China during 2018 and 2019 (Table 3).The results showed that DPEN and DPRN could obtain higher grain yield,NUE and RUE,while no significance difference existed for either grain yield or RUE,regardless of seasons and years. In addition,NUE in DPRN was increased significantly by 19.10–27.97% in 2018,and by 14.99–23.15% in 2019 relative to that in DPEN. Basically,the tendencies were similar with those seen in the results obtained at Guiping (Tables 2 and 3; Fig.6).

4.Discussion

4.1.Crop performances based on agronomic and physiological traits in optimized cultivation managements

Many studies have shown that greater biomass production,longer LAD,larger LAI,and higher productive tiller percentages were important parameters for ultimately obtaining a higher yield (Linet al.2009; Wanget al.2014;Huanget al.2016). The cultivation management practices with respect to planting density and N rate have been studied and the results indicate that they can increase both grain yield and NUE due to the improvement of various agronomic and physiological traits of the rice (Zhanget al.2018,2019;Panet al.2019; Xieet al.2019),especially increasing the biomass and N accumulation after flowering (Huanget al.2019). The present study suggests that DPEN and DPRN could not only significantly increase the grain yield,NUE and RUE (Table 2; Fig.6),but also improve the agronomic and physiological characteristics of the rice population compared to FP conditions. Dry matter accumulation,CGR(Fig.3),LAD (Fig.4),N uptake (Fig.5),and RUE (Fig.6)were increased significantly during the period from PI to HD,while panicle number,sink size (Table 2),and apparent transformation rate of dry matter (Fig.4) were enhanced dramatically. Correlation analysis indicated that panicle number m–2was positively correlated with grain yield,while no correlation was found between grain yield and either spikelets per panicle,seed set rate or 1 000-grain weight(Fig.7),implying that panicle number m–2is important for achieving higher grain yield. In addition,grain yield was significantly and positively correlated with total dry weight,while no correlation was found between grain yield and HI,implying that dry weight accumulation is an important factor in determining grain yield,while HI is not. It is noteworthy that LAD and CGR from PI to HD,leaf N concentration at HD,and apparent transformation ratio,NUE and RUE were positively correlated with grain yield (Fig.7). Together,these findings indicate that the growth phase from PI to HD is a key stage for rice production in the double-cropping rice system of South China,and the higher grain yield and resource use efficiencies mainly resulted from higher dry matter accumulation,CGR,LAD,N uptake,and RUE between PI and HD,as well as from higher panicle number,sink size,and apparent transformation rate of dry matter through the synchronous optimization of both planting density and N application strategies.

Fig.6 The change in radiation use efficiency (RUE) at each rice growth phase (A and B) and the partial factor productivity of applied N fertilizer (PFPN) (C and D) under various treatments at Guiping site,Guangxi,China during 2019 early (A and C) and late (B and D) seasons. FP,farmer’s practice; DPEN,dense planting with equal N input and delayed N application; DPRN,dense planting with reduced N input and delayed N application. RUE was calculated as the ratio of aboveground total dry weight to accumulated intercepted radiation over the period of each growth stage. TP,transplanting; MT,mid tillering; PI,panicle initiation;HD,heading stage; MA,maturity; WGP,the whole growth period of rice. Bars are mean±SE (n=3). Different letters indicate statistical significance at the P=0.05 level within the same growth phase and the same season.

4.2.lnfluences of planting density and N management on tillering and panicle number

Fig.7 Correlations between grain yield and panicle number m–2 (A),spikelets per panicle (B),seed set rate (C),1 000-grain weight(D),dry weight of whole growth period (E),harvest index (HI) (F),leaf area duration (LAD) from panicle initiation (PI) to heading stage (HD) (G),crop growth rate (CGR) between PI and HD (H),apparent transformation ratio (I),leaf nitrogen concentration at HD(J),the partial factor productivity of applied N fertilizer (PFPN) (K),and radiation use efficiency (RUE) of the whole growth period(L) in the early and late seasons of 2019.

An appropriate increase in planting density can reportedly increase the panicle number,and ultimately produce a higher grain yield (Huanget al.2013; Wang and Peng 2017).Huanget al.(2018) found that increasing planting density can compensate for the yield loss from N input reduction.Therefore,an appropriate increase of planting density is of great importance for augmenting panicle number and grain yield. The results of the present study showed that when compared with the FP,DPEN and DPRN could significantly increase panicle number by 15.8–26.8% and 26.1–34.9%,respectively,and higher grain yield (Table 2)could be obtained by increasing planting density from 18.5 to 28 hills m–2. However,there were significant differences in the ways that greater panicle numbers were obtained in DPEN and DPRN,i.e.,DPEN relied on increasing the maximum tiller number at PI,while DPRN raised the productive tiller percentage (Fig.2). The higher N rate in DPEN (21.4% higher than DPRN) applied at basal and mid tillering stage led to higher tiller numbers at the early growth stage (Table 1; Fig.2),and more dead tillers due to limited space and nutrition in the late growth stages.This resulted in a reduced productive tiller percentage and ultimately produced a panicle number that was comparable with DPRN (Fig.5),which was consistent with the results previously reported (Penget al.2011; Lianget al.2019; Liu Het al.2019). This finding implied that increasing N input cannot improve the productive tiller percentage under a higher planting density condition. Although a lower N rate(150 kg ha–1,23.1% lower than FP and DPEN) was used,DPRN with proper N application at PI resulted in a higher productive tiller percentage (Table 1; Fig.2). This result is consistent with a previous study which found that optimizing the total N input and adjusting the topdressing time could have positive effects on rice yield and NUE (Penget al.2010; Zhonget al.2010; Xueet al.2013,Liu Het al.2019).

It is proposed that the supplied N rate influences the development of tiller primordia,and the uptake of N may bea good indicator of the tillering activity (Tanaka and Garcia 1965). Previous studies found that a higher N uptake was positively correlated with panicle number (Huanget al.2011;Zhuet al.2015; Liu Yet al.2019). These studies indicated that N uptake is more important in determining tiller number than the N rate supplied. In the present study,the leaf N concentrations at HD in DPEN and DPRN were significantly higher than that in FP,and the leaf N concentration at HD in DPRN was lower than that in DPEN (Fig.5). However,DPRN could increase the productive tiller percentage more significantly than DPEN (Fig.2). These observations indicated that the moderate leaf N concentration in DPRN produced a higher productive tiller percentage,while the excess leaf N concentration in DPEN generated a large number of unproductive tillers (Figs.2 and 5). Therefore,although the N rate supplied in DPRN was 23.1% lower than those in FP and DPEN (Table 1),the proper timing of split application in DPRN significantly increased leaf N concentration at HD (Fig.5),which promoted an increase in the productive tiller percentage (Fig.2) and ensured a considerable panicle number and ultimately resulted in a higher grain yield than that in FP (Table 2).

Table 3 The grain yield,NUE and RUE of the whole growth period at Xingning site,Guangdong Province,China in the early and late seasons during 2018 and 2019

4.3.The influence of dry matter translocation on grain yield

Whether grain filling mainly relies on assimilates produced after flowering and/or assimilates redistributed from the reserves before HD remains debatable (Yang and Zhang 2010). Panet al.(2020) showed that the improvement of yield was due to a higher above-ground biomass before heading,while Liu Het al.(2019) reported that maintaining daily biomass production could alleviate yield loss during the grain-filling stage. The present study observed that grain filling in DPEN and DPRN mainly depended on assimilates that were redistributed from the reserves in stems and/or leaves before HD,which exhibited a significantly higher apparent transformation rate of dry matter than that in FP(Fig.4). Moreover,greater dry matter accumulation and CGR during the growth phase from PI to HD were observed,implying that the growth phase from PI to HD is an important stage for the assimilates to be redistributed to grain filling(Fig.3) in the double-cropping rice system of South China.This is consistent with the results reported previously by Xieet al.(2019). In addition,compared to FP and DPEN,the dry matter accumulation,CGR,RUE,and N uptake from HD to MA were increased significantly in DPRN,and meanwhile the productive tiller percentage,sink size and spikelets to leaf-area ratio were improved significantly in DPRN in the late season (Figs.2,4–6).

4.4.The influence of different cultivation management strategies on RUE

Zhanget al.(2015) considered solar radiation to be the main factor limiting rice growth and grain yield.Many previous studies found that a higher RUE is associated with higher incident solar radiation,higher intercepted solar radiation by the rice canopy,upright top leaves and smaller leaf angles,stronger leaf photosynthesis,weaker respiration,and other factors (Huanget al.2016; Hubbartet al.2018;Chenet al.2019). The present study showed that RUE in DPRN reached 1.60 and 1.80 g MJ–1,which were 28.6 and 37.9% higher than the RUE in FP in the early and late seasons,respectively. DPEN and DPRN could significantly promote the RUE in the growth phase from PI to HD,and consequently enhance RUE of the whole growth duration,compared to FP (Fig.5). This tendency is similar with those of dry matter accumulation,CGR,LAD,and N uptake (Figs.3 and 6). In addition,correlation analysis showed that the RUE was positively correlated with grain yield regardless of the seasons and treatments (Fig.7). This indicated that greater RUE would result from the improved agronomic and physiological traits of the rice population,and ultimately enhance the grain yield. Moreover,N fertilizer application regimes could improve RUE by increasing LAD,allocating more N to leaves,and leading to the interception of more solar radiation (Druilleet al.2019; Heet al.2019). Liet al.(2012) reported that an optimal N rate is helpful to achieve a higher light interception rate,RUE and HI. In the present study,with a 51.4% higher planting density and 23.1% lower applied N fertilizer,DPRN significantly increased N uptake,tiller number from PI to HD,and dramatically improved leaf N concentration at HD and LAD from PI to HD in comparison with FP (Table 1; Figs.2 and 5). These results indicated that the higher LAD and leaf N concentration in DPRN contributed to the higher RUE (Fig.6).

4.5.The differences in grain yield and NUE between DPEN and DPRN

It should be noted that there was no difference in grain yield between DPEN and DPRN (Tables 2 and 3),but NUE in DPRN was significantly greater than that in DPEN. This indicated that,under a greater planting density condition,further increases in the N rate and topdressing fertilizer N at the booting stage (increasing fertilization times) could not significantly increase grain yield and NUE in the double cropping rice system of South China. The results of two years of comparison trials further revealed that DPEN and DPRN enhanced grain yield and RUE,while DPRN had a higher NUE than DPEN (Table 3). In addition,the present study found that annual rice production in DPRN was very high and reached 20.38 t ha–1in the same paddy field at the Guiping site in 2019,which attained 88.3% of the highest grain yield on record (23.07 t ha–1) in a different paddy field in South China (Zhouet al.2017). Therefore,with reduced N input and proper N topdressing at PI,DPRN is a feasible approach to simultaneously promote both grain yield and resource use efficiencies in the double rice cropping system of South China.

5.Conclusion

Significant differences in grain yield and nitrogen and radiation use efficiencies were found among FP,DPEN and DPRN treatments. Compared to FP,DPRN had a greater planting density,reduced N input and delayed N application,and it achieved significantly higher grain yield,NUE and RUE. However,DPEN,with an equal planting density,further increased N input and one additional N application in comparison with DPRN,had a grain yield and RUE comparable to DPRN,but its NUE was significantly lower than that in DPRN. Increased CGR,LAD,N uptake from PI to HD,a higher productive tiller percentage and a greater apparent transformation ratio of dry weight from stems and sheaths to panicles were critical parameters for the simultaneous improvements in grain yield,NUE and RUE in DPRN. We conclude that DPRN is a feasible approach to promote synchronized improvements in grain yield and nitrogen and radiation use efficiencies in the double rice cropping system in South China.

Acknowledgements

This work was supported by the National Key Research and Development Program of China (2016YFD0300108-5),the Natural Science Foundation of Guangdong Province,China(2017A030313110,2018A030313463),the Discipline Team Building Project of Guangdong Academy of Agricultural Sciences,China (201617TD),the Special Fund for Scientific Innovation Strategy,China (Construction of High-Level Academy of Agricultural Science),and the Guangdong Provincial Key Laboratory of Applied Botany,South China Botanical Garden,Chinese Academy of Sciences(AB2018013).