Huan Chen,Aixing Deng,Weijian Zhang,Wei Li,Yuqiang Qiao,Taiming Yang,Chengyan Zheng,Chengfu Cao,*,Fu Chen*

aCollege of Agronomy and Biotechnology,China Agricultural University,Beijing 100193,China

bInstitute of Crop Sciences,Chinese Academy of Agricultural Sciences,Beijing 100081,China

cCrop Research Institute,Anhui Academy of Agricultural Sciences,Hefei 230031,Anhui,China

d Anhui Center of Agricultural Meteorology,Hefei 230031,Anhui,China

Keywords:Winter wheat Grain yield Yield stability AMMI analysis Long-term fertilization

ABSTRACT An understanding of wheat yield and yield stability response to fertilization is important for sustainable wheat production.A 36-year long-term fertilization experiment was employed to evaluate the yield and yield stability of winter wheat.Five fertilization regimes were compared,including(1)CK,no fertilizer;(2)NPK,inorganic fertilizer only;(3)O,organic fertilizer only;(4)NPKO,50%of NPK plus 50%of O,and(5)HNPKO,80%of NPK plus 80%of O.The greatest yield increase was recorded in HNPKO,followed by NPKO,with O producing the lowest mean yield increase.Over the 36 years,the rate of wheat yield increase in fertilized plots ranged from 95.31 kg ha-1year-1in the HNPKO to 138.65 kg ha-1year-1in the O.Yield stability analysis using the additive main effects and multiplicative interactions(AMMI)method assigned 62.3%,26.3%,and 11.4%of sums of squares to fertilization effect,environmental effect,and fertilization×environment interaction effect,respectively.The combination of inorganic and organic fertilization(NPKO and HNPKO)appeared to produce more stable yields than O or NPK,with lower coefficients of variation and AMMI stability value.However,wheat grown with O seemed to be the most susceptible to climate change and the least productive among the fertilized plots.Significant correlations of grain yield with soil properties and with mean air temperature were observed.Thesefindingssuggestthatinorganic+organic fertilizercanincreasewheatyield and its stability by improvement in soil fertility and reduction in variability to climate change.

1.Introduction

Wheat(Triticum aestivum L.)is the primary cereal crop in China,accounting for about 15%of the world's wheat grain production[1].Although large increases in wheat yield have been achieved during the past century by improvement of agronomic practices and cultivar selection[2],a marked yield stagnation has been reported and the rate of yield increase has slowed to 0.9%per annum[3,4].Increasing and sustaining wheat yield is vital for meeting the growing food demand.

Yields in crop production systems are affected by many factors,including genotype,agronomic management,and weather[5–9].Genetic improvement contributed dramatically toyieldincreasesincenineteenfifties,andgeneticgraininyield varied in different investigations[4].As to agronomic management,its overall impacts on crop yield have been well investigated,buttheyieldadvantageordisadvantageofspecific management practices may be obscured by evaluation of only main effects.Temporal variation in crop yields reflects the stability of anagriculturalsystem inresponse to environmental change,with a stable system being the one that varies least in response to ambient variation[10].Measurement of crop yield stability isimportant to sustainable agriculture,given that large fluctuations in crop yields are predicted to accompany global climate change[11].Thus,both yield and yield stability should be included in crop yield estimation[12].

Stability analysisofcrop yieldsrevealsyear-to-year variability in the form of year×treatment interaction effect,a statistic that yields more information than average yield[13].However,it is difficult to explain interactions by a conventional analysis of variance approach.Statistical approaches commonly used in yield stability evaluation of crop genotypes[14,15]have been applied for yield comparison across different agronomic systems[6,7,16–18].The coefficient of variation(CV)is popularly employed to quantify yield stability[19],and regression stability analysis[17]is another effective technique for identifying year×treatment interaction,owing to its simplicity and illustrative appeal.The additive main effects and multiplicative interaction(AMMI)model is an extended regression model to which an environmental main effect has been added along with more than one multiplicative term.It is regarded[20]as a preferred method for interaction description and simultaneous selection for yield and yield stability in agricultural systems.

Long-term experiments have been widely established in agricultural studies,and have provided valuable data that has not only minimized unknown effects through multi-year trials[21]but also played an irreplaceable role in validating existing models and predictions[22,23].Likewise,long-term fertilization field trials have been established for investigating crop yield and yield stability.Ma et al.[24]found that balanced mineral fertilization could increase and stabilize corn yield in northeast China,but Bhattacharyya et al.[25]found a decreasingtrend ofcrop yield underNPK fertilization.Bilsborrow et al.[2]observed that conventional management outyieldedorganicmanagement.Noadditionalpositive effectsofpigslurryplusstraw togetherwith mineral fertilizers on wheat yield were found[26].However,Qin et al.reported that NPK plus manure increased crop yields more than NPK alone or NPK plus straw[5].

Although such long-term trials can help identify agricultural systems with great productivity and high stability,limited information is obtainable on the yield stability of wheat in response to different fertilization regimes in the Huanghuai region in China,which contains＞15 Mha of wheat cultivation area and is responsible for two thirds of Chinese wheat production[27].We initiated in 1981 a continuous long-term fertilization experiment to assess(1)grain yield,yield trends,and yield stability under different fertilization regimes(2)climate change during the experimental period and its relationship with yield,and(3)the dynamics of soil properties under different fertilization regimes and their relationship with yield.

2.Materials and methods

2.1.Site description

A long-term field fertilization experiment was initiated in 1981 at Yangliu Agricultural Station(116°45′E,33°37′N(xiāo)),Suixi County,Huaibei city,Anhui province,China.Wheat–maize rotation is a major cropping system in this region.The average annual precipitation was 807.0 mm and the average annual temperature 15.0°C from 1981 to 2016(Fig.1,data from Suixi Meteorological Station).

The soil is classified as a vertisol,according to World Reference Base soil taxonomy[28],and developed on fluviolacustrine deposits.The initial topsoil properties before the experiment were asfollows:pH 7.6,organic matter 10.22 g kg-1,total nitrogen 0.78 g kg-1,total phosphorus 0.47 g kg-1,availablenitrogen64.10 mg kg-1,availablephosphorus2.50 mg kg-1,and available potassium 118.20 mg kg-1.

2.2.Experimental design

Thelong-termexperimentinvolvedfivetreatments,eachapplied tofourplotsof30 m2:(1)CK,nofertilizeradded;(2)NPK,inorganic fertilizer only(N 525 kg-1ha-1a-1,P2O5210 kg-1ha-1year-1,K2O 210 kg-1ha-1year-1);(3)O,organic fertilizer only(7500 kg ha-1year-1,equivalenttoN525 kg-1ha-1year-1);(4)NPKO,50%ofNPK plus 50%of O;(5)HNPKO,high application ratio of inorganic and organicfertilizers,80%ofNPKplus80%ofO.AllPandKfertilizers were applied before wheat or maize sowing as basal fertilizers.Inorganic N fertilizer was applied basally in portions of 60%and 45%to wheat and maize,respectively,with the remainders appliedastopdressing at the jointing phase ofwheat and the bell stage of maize.The inorganic fertilizer comprised urea,superphosphate,and potassium chloride.The organic fertilizer was soybean cake compost containing 300–400 g kg-1organic carbon,60–70 g kg-1total nitrogen,10–30 g kg-1P2O5,20–30 g kg-1K2O,and 100–150 g kg-1water.

The field was plowed to a depth of 15–20 cm after maize harvest in every year.Winter wheat was planted at a density of 225×104plants per hectare and a spacing of 20 cm in early October,and harvested in early June.Varieties of winter wheat used in the experiment were those prevalent in the region in each period(Table 1).Pesticides,fungicides,and herbicides were applied as needed during the growing season.Irrigation was usually applied depending on crop growth and at sowing and fertilization.After harvest,aboveground straw was completely removed from each plot.

2.3.Data collection

The wheat from each plot was manually harvested at maturity,and the grain was air-dried and the weight corrected to 12%moisture content.Final grain yield of wheat in each plot was monitored from 1981 to 2016,and yield for each treatment in each year was calculated as the mean value of the replicates and expressed in kg ha-1.Monthly weather data during the study period,including average temperature(°C),precipitation(mm),and sunshine duration(h)were obtained from the Suixi meteorological station.

Fig.1–Average annual air temperature and annual precipitation from 1981 to 2016.

2.4.Soil sampling and analysis

In every year starting in 1981,except the decade of the 2000s,topsoil samples from 0 to 20 cm depth in each plot were collected as a mixture of 5–10 subsamples after the summer maize harvest and before winter wheat sowing.Soil samples were air dried to stable water content and ground for chemical analysis[29].Soil organic matter(SOM)was analyzed by potassium dichromate oxidation-redox titration,total nitrogen(TN)was determined by Kjeldahl digestion–distillation,availablenitrogen(AN)wasmeasured by thealkaline permanganate method,available phosphorus(AP)was assayed by the Olsen method,and available potassium(AK)was extracted with ammoniumacetateandthen determinedbyflamephotometry.

2.5.Statistical analysis

Effects of fertilization treatments on grain yields of winter wheat were analyzed by one-way ANOVA,and the significance was determined by Duncan's test.The Pearson correlation coefficient was used to measure the strength of a linear association between meteorological indices,soil properties and wheat yield,using a two-tailed test.

Table 1–Varieties of winter wheat cultivated during each period.

Trends of wheat yield and various soil parameters under the different regimes over 36 years were determined by linear(least squares)regression,using the equation

where Y is the grain yield of winter wheat or content of a soil parameter,α is a constant,β is the regression coefficient representing the trend(for yield:kg ha-1year-1;for soil parameters:g kg-1year-1or mg kg-1year-1),and t is the year of the experiment.The P-value from t-test indicates the significance of the β.

The variability of yield was expressed as the coefficient of variation(CV)of yield over 36 years.

Yield stability of winter wheat was estimated by the AMMI model,using the formula

where Yijis the arithmetic mean yield of wheat of the ith fertilization treatment in the jth year,μ is the overall mean,Fiis fertilizer main effect,Ejis environmental main effect,n is the number of(interaction principal component axis)IPCA axes,λnisthesingular valueforthe IPCA,γinand δjnare the eigenvectors of fertilization and environment respectively,εijis the residual.

The AMMI stability value(ASV)was calculated[15,24,30]as

For statistical analysis,Microsoft Excel 2010(Microsoft Corporation)was used for basic calculation.ANOVA was determined by IBM SPSS Statistics 22(International Business Machines Corporation).AMMI was analyzed by Data Processing System(DPS)17.0(Hangzhou Ruifeng Information Technology Corporation Limited).Figures were drawn with Origin 8.0(OriginLab Corporation).

Fig.2–Mean grain yield under different fertilization regimes.CK,no fertilizer;NPK,inorganic fertilizer only;O,organic fertilizer only;NPKO,50%of NPK plus 50%of O;HNPKO,80%of NPK plus 80%of O.Different lowercase letters indicate significant differences between treatments(P＜0.05).

3.Results

3.1.Yield and yield trends

The mean yield of wheat across 36 years differed significantly among treatments(Fig.2),ranging from 709.95 kg ha-1for CK to 5710.67 kg ha-1for HNPKO.Application of fertilizers significantly increased yield(P＜0.05),resulting in an increase from 544.2%to 704.4%relative to CK.The ranking of the increase in the fertilizer treatments,from high to low yield,was HNPKO＞NPKO＞NPK＞O.No significant difference was observed between O and NPK.Organic plus inorganic fertilizer(NPKO and HNPKO)markedly increased the mean yield of wheat relative to O,whereas no significant difference was observed between NPKO and NPK.

Owing to different fertilization and weather change,wheat yields exhibited diverse trends in different treatments(Fig.3).According to the linear regression,grain yield trends ranged from-7.75 kg ha-1year-1to 138.65 kg ha-1year-1and were lowest for CK and highest for O.Positive(P＜0.001)yield trends were observed in the fertilizer treatments.

3.2.Yield stability

Fig.3–Linear regressions of wheat yield upon years under different fertilization regimes from 1981 to 2016.CK,no fertilizer;NPK,inorganic fertilizer only;O,organic fertilizer only;NPKO,50%of NPK plus 50%of O;HNPKO,80%of NPK plus 80%of O.

Table 2–Analysis of variance and AMMI analysis of winter wheat yields(t ha-1).

Fertilization was treated as one of the main effects and weather in each year as the other main effect,“environment”.According to the AMMI analysis,61.8%,26.9%,and 11.3%of the sum of squares were attributed to fertilization effect,environmental effect,and fertilization×environment interaction effect(Table 2),indicating that the effect of fertilization×environment interaction could not be ignored in yield stability estimation.Moreover,both IPCA1 and IPCA2 were significant,and contributed 83.4%to the sum of squares for the F×E interaction.

Results of AMMI analysis were interpreted using biplots.In the biplot of IPCA1 and mean yield(Fig.4),the points for fertilization treatments at the right hand side of the mean value(x=4276.79)and close to the line PCA1=0 indicated high mean yields and negligible fertilization×environment interaction.Thus,it was clear that the three treatments NPK,NPKO,and HNPKO displayed general adaptability overall years.However,O was suited only to the favorable years of weather,as indicated by high yield and large PCA1 scores.

A biplot of IPCA1 and IPCA2 for environment is shown in Fig.5.The distance from a point to the origin corresponds to the yield stability:the shorter the distance,the higher the yield stability.HNPKO and NPKO are located around the center of the biplot and may thus be considered stable CK displays the lowest stability by its long distance from the origin,and NPK and O display medium levels of yield stability,with NPK more stable than O.

The CV and ASV were calculated to quantify the yield variability and stability(Table 3).The CV corresponded well with the ASV,showing the highest value in CK and the lowest in HNPKO.

3.3.Meteorological variation and its correlation with yield

The meteorological factors changed over 36 years.The annual and wheat seasonal mean temperature increased significantly by 0.28 °C and 0.32 °C,respectively,per 10 years(Fig.6a).Average annual precipitation at the station was about 800 mm,of which more than two thirds fell in June–September owing to the subtropical–warm temperate transitional monsoon climate and the other third fell in the growing season for wheat.Neither annual precipitation nor wheat seasonal precipitation changed significantly from 1981 to 2016,but April precipitation showed a significant increasing tendency(Fig.6b).Similarly,there was no significant variation in annual and wheat seasonal sunshine duration,whereas the sunshine duration in March increased over the 36 years(Fig.6c).

Fig.4–Bioplot of IPCA1 and mean yield of winter wheat.CK,no fertilizer;NPK,inorganic fertilizer only;O,organic fertilizer only;NPKO,50%of NPK plus 50%of O;HNPKO,80%of NPK plus 80%of O.e1,e2,e3,…,e35,and e36 indicate environments of years from the start of the experiment.

Fig.5–Bioplot of IPCA1 and IPCA2.CK,no fertilizer;NPK,inorganic fertilizer only;O,organic fertilizer only;NPKO,50%of NPK plus 50%of O;HNPKO,80%of NPK plus 80%of O.e1,e2,e3,…,e35,and e36 indicate environments of years from the start of the experiment.

According to the correlation analysis(Table 4),both annual and wheat seasonal mean temperature were correlated significantly with wheat yield(P＜0.05),whereas neither precipitation nor sunshine duration was correlated with grain yield(P＞0.05).

3.4.Soil properties and their correlation with yield

Fertilization influenced soil variables(Fig.7).SOM increased significantly for all treatments(P＜0.001)over 36 years,compared with the initial value of 10.22 g kg-1.The slope of the SOM trend line under different fertilization regimes ranged from 0.02 g kg-1year-1to 0.46 g kg-1year-1,and the largest slope occurred in NPKO(Fig.7a).Similarly,soil TNcontent increased statistically during the period of study in all treatments.The TN growing rate for CK,NPK,O,NPKO,and HNPKO were 7.24,14.64,18.91,16.22,and 18.86 mg kg-1year-1,respectively(Fig.7b).AN content was increased significantly in all plots except CK(Fig.7c),and the rates of AN increase with organic amendment were higher than those with mineral fertilization only.The AP content in soil under nonfertilization showed a slight decline(P＞0.05).The AP trend lines showed pronounced rising tendencies in the treatments with inorganic addition(NPK,NPKO,and HNPKO),whereas there was no increase of AP in O(P＞0.05).In contrast,the AK content showed declining trends(slope＜0)under all regimes,but significantly so only in CK and HNPKO(P＜0.05).

Table 3–Variability and stability parameters of wheat yield under different fertilization regimes.

Correlation analysis showed that grain yield of winter wheat was positively and significantly correlated with soil parameters excluding AK(P＜0.05,Table 5).

4.Discussion

Yield and yield stability of winter wheat in the Huang-Huai-Hai plain of China were investigated in a long-term fertilization experimentbegun in1981.Under continuousand different fertilizer applications,average yields of wheat across the 36 years differed significantly and the mean values under NPK,O,NPKO,and HNPKO were respectively 6.0,5.4,6.6,and 7.0-fold greater than those under CK(Fig.2).This finding indicated that greater yields were produced by a combination of mineral fertilizer and organic amendment,in accord with previous studies[5,31–33].The greater increase of wheat yield in inorganic plus organic fertilized plots might be ascribed to the increase of soil organic matter and the increase of soil nutrient availability,a conclusion supported by the finding that the combination of inorganic and organic fertilization yielded higher rates of increase of SOM,TN,and AP than did NPK or O(Fig.7).Similar conclusions have been reported previously[5,33–35].

Fig.6–Linear regressions of air temperature,precipitation and sunshine duration upon years from 1981 to 2016.

Distinct yield dynamic changes were found under different fertilization regimes(Fig.3).Wheat yield with no fertilizer(CK)declined slightly(P＞0.05)and remained at the lowest level,though new wheat cultivars with higher yield performance than old cultivars were introduced[36,37].The slight yield depression in non-fertilized plots was probably associated with depletion of available nutrients[34],as available potassium concentration in our results(Fig.7e),and an accompanying imbalance of nutrients needed for wheat growth[5].Also,climate change may have exerted adverse impacts on yield [38].In contrast,yields increased substantially in fertilized plots,with the annual yield growth(slope)ranging from 95.31 kg ha-1for HNPKO to 138.65 kg ha-1for O.Longterm fertilization led to increased soil organic matter and improved nutrient availability over time(Fig.7),also contributing to increased wheat growth and development.This conclusion is supported by the significant and positive correlations between yield and soil properties(Table 5).Notably,a high application rate of inorganic plus organic fertilizer(HNPKO)was disadvantageous to yield growth rate despite its high mean yield performance(Fig.1).Zhang et al.[33]also found that high application rate of manure only or of total N fertilizer reduced the mean annual change of wheat yield over time in long-term experiments in North China.In the present study,O displayed a greater yield growth rate despite a lower mean yield than other fertilization regimes(Fig.3),implying that O had the most potential to enhance wheat yield.Likewise,Cai and Qin[39]and Yan and Gong[40]found that wheat yield under solely organic amendment displayed a higher increasing trend than solely inorganicaddition and combination ofinorganic and organic application.

Table 4–Correlation between wheat yield and meteorological variables.

Fig.7–Linear regressions of soil properties upon years under different fertilization regimes.(a)Soil organic matter,(b)total nitrogen,(c)available nitrogen,(d)available phosphorus,and(e)available potassium.CK,no fertilizer;NPK,inorganic fertilizer only;O,organic fertilizer only;NPKO,50%of NPK plus 50%of O;HNPKO,80%of NPK plus 80%of O.

Table5–Correlation between wheatyield and soil properties.

Many factors have been reported to affect the grain yield of wheat,including fertilization,cropping system,precipitation,and crop variety[23,41,42].In the present study,we focused on fertilization,climate variables,and their interaction.The fertilization,environment(weather change)and fertilization×environmentinteraction contributed respectively 61.8%,26.9%,and 11.3%of the sum of squares(Table 2),confirming the significant effects of meteorological variation and their interaction,aswellasoffertilization,on wheatyield variation.Compared with the yield trend,the fluctuation of wheat yield from one year to the next reflected yearly weather variation.Yield variability and stability differed among fertilization regimes,according to the CV and ASV(Table 3),showingthatthecombinationofmineralandorganic fertilization appeared to be more stable than solely mineral or organic amendment.Pan et al.[43]reported that average yield variability was very significantly and negatively related to cropland SOM level,such that a 1%increase in SOM would lead to a 3.5%decrease of yield variability.In contrast,O tended to be more dependent on climate change than other fertilized treatments,regardless of its higher level of soil organic matter(Fig.7).This dependence might be attributed to the influence of temperature on soil nutrient release from organic amendments[44],and the unfavorable effects would be mitigated by mineral fertilizer addition.

There is compelling evidence that increases in crop yield were associated with soil organic carbon stock[35,40,43],nutrient content[5]and meteorological factors[5,26,39].Similarly,the SOM increasing trend was in good agreement with the annual yield change(Table 5)in the present study.Although positive and significant correlations between grain yield and soil TN,AN,and AP were observed(Table 5),AK was not significantlycorrelatedwith wheat yield.Furthermore,only mean temperature,among the meteorological indices,showed a significant relationship with yield,whereas Zhang et al.[23]found that sunshine duration,relative humidity,and wind speed were also substantially correlated with wheat yield.Although wheat yield was insignificantly associated with precipitation and sunshine duration in our study,it should not be ignored that a significant increase of sunshine hours in March and precipitation in April(Fig.6b,c)might influence wheat growth at the elongation and booting stages.

In addition to long-term experiments applying constant and unchangeable treatments for years and providing more valuable information than short-term studies[21–23],extensive modifications have been developed,as crop varieties and plant protection practices,to ensure the continuity and relevance of the long-term experiments[45,46].In the present study,we changed the wheat varieties 12 times during the 36 years(Table 1),and introduced pesticides,fungicides,and herbicides that were in general use in specific decades.These factors,especially cultivar renewal,are intrinsic to crop yield increase and stabilization[7,8,11,41].Varietal improvement of winter wheat has been found to increase yield by 0.48%–1.23%annually in northern China[47].Because the contributions of genetic improvement and its interaction with soil nutrient trends and climate change could not be evaluated in the present study,additional independent trials or calibrated crop models will be needed in further research.

5.Conclusions

On average,grain yield of winter wheat was significantly increased by fertilization.The greatest increase relative to CK was found under the HNPKO treatment,followed by NPKO,while O displayed the lowest increase.Increasing trends of grain yield since 1981 were found in fertilized plots,while a slightly decreasing trend was observed in CK.According to AMMI analysis,wheat yield stability was markedly increased by fertilization,among which combinations of inorganic and organic fertilization(NPKO and HNPKO)appeared to be the most stable,with lower coefficients of variation and AMMI stability value.O appeared more susceptible to weather variation,despite its higher yield increase rate.Wheat yield was correlated significantly and positively with mean air temperature and soil properties.In summary,the yield of winter wheat tended to be higher and more stable under a combination of inorganic and organic fertilization,a result attributed to improvement in soil fertility and reduction in variability to climate change.

Acknowledgments

We thank the staffofSuixiAgriculturalStation for conducting and managing the long-term field experiment.This work was supported by the National Key Research and DevelopmentProgram ofChina(2016YFD0300803),the Special Fund for Agro-scientific Research in the Public Interest(201503116-10),the Agricultural Science and Technology Innovation Program(CAAS-XTCX2016019-03 and Y2016XT01-03),and the Science and Technology Major Project of Anhui Province(16030701099).