Wenji Yng,Weijin Liu,Yulin Li,Shiwen Wng,c,d,*,Lin Yin,c,d,Xiping Deng,c,d

a State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau,College of Natural Resources and Environment,Northwest A&F University,Yangling 712100,Shaanxi,China

b College of Life Sciences,Northwest A&F University,Yangling 712100,Shaanxi,China

c Institute of Soil and Water Conservation,Northwest A&F University,Yangling 712100,Shaanxi,China

d Institute of Soil and Water Conservation,Chinese Academy of Sciences & Ministry of Water Resources,Yangling 712100,Shaanxi,China

Keywords:

A B S T R A C T Erratic rainfall and misalignment between the rainy season and the growing season of winter wheat greatly limit rainfed winter wheat yield in the Loess Plateau of China.To increase the grain yield of winter wheat in this region,the effects of different agronomic practices,including adjusting planting pattern(NR,narrow row spacing),increasing seeding rate(high seeding rate,HS),decreasing basal nitrogen rate and increasing top-dressed nitrogen rate(DBN),and replacing an old cultivar with a new cultivar(NC)on wheat yield were investigated for two consecutive years.The results showed that the current grain yield of rainfed winter wheat in the Loess Plateau could be increased to 5879–7093 kg ha-1 by HS,DBN and NC practices relative to the practice of high-yielding farmers(PF).The increased yield due to HS,DBN and NC was attributed to the higher number of spikes ha-1,1000-grain weight,and kernels spike-1.Before the flowering stage,HS increased soil water consumption(SWC)in 1–3 m subsoil due to the higher plant population compared with that of PF,whereas DBN decreased SWC in the 0–2 m soil layer compared with that of PF.After the flowering stage,HS,DBN,and NC increased SWC by 8–16 mm in 2–3 m subsoil compared to PF.The water use efficiency(WUE)was increased under DBN and NC in comparison with PF.However,the WUE did not increase under HS as it had the highest evapotranspiration among the five treatments.Increasing the use of subsoil water during the late growth stage by optimizing agronomic practices or applying new cultivars with expansive roots should be the primary approach to increase rainfed winter wheat yield in this region.

1.Introduction

Rainfed agriculture is the primary farming practice in 75%of the global croplands and is widely applied in drylands,where crop yield is frequently threatened by unfavorable climate conditions[1].In China,the Loess Plateau is a typical dryland area with 16 million ha of cropland.Here,44%of the croplands are planted with rainfed winter wheat[2,3],and the wheat yield is consistently limited by erratic and low annual rainfall[4,5].However,with the improvement of agronomic practices and alternation of cultivars,the wheat yield and water use efficiency(WUE,grain yield per unit crop evapotranspiration)in the Loess Plateau has steadily increased over the past 60 years.For instance,the average yield and WUE during 1972–1993,were 3745 kg ha-1and 10.24 kg ha-1mm-1,respectively,and from 1994 to present,the average yield and WUE increased to 4357 kg ha-1and 12.19 kg ha-1mm-1,respectively[6–12].Nevertheless,with increases in population and frequency of unfavorable weather,food safety and quality remain key issues in China.Thus,it is essential to continuously improve winter wheat yield and WUE in this area.

The rainy season and growing season of winter wheat do not coincide in this area,making the soil available water crucial for wheat production.At the same time,the soil of the Loess Plateau is an excellent water reservoir because it possesses a very deep loessial soil layer and can hold up to 800 mm of rainfall to a depth of 3 m[6,13].Therefore,most rainfall,especially during the fallow period,can be stored in the soil.Previous study has shown that approximately 80% of the evapotranspiration(ET)occurring in rainfed winter wheat came from water stored in the soil at sowing during dry years,reducing to 50%–60% during wet years[14].In other words,if soil water consumption increases,the yield and WUE of rainfed winter wheat would substantially increase in this area,especially in dry years[15–17].However,even in highyielding fields,a large amount of soil water that can be used by wheat remained in the 0–3 m soil layer at harvesting,especially in 1–3 m subsoil[2,15,16,18].Therefore,increasing the use of soil water,particularly subsoil water,could assist in continuously improving wheat yields in this area.

Agronomic practices and cultivars generally contribute the most to the improvement of crop yield.Agronomic practices such as planting pattern,seeding rate,and nitrogen(N)application strategy,as well as cultivars,influence the pattern of use of soil water[19–22].For instance,narrow row spacing with a larger area of crop canopy reduces soil evaporation in the early stage[23,24].Increased seeding rate enhances the intraspecific competition of winter wheat for nutrients and water resulting in increased root length density of winter wheat in the subsoil[22,25].Decreasing basal N rate properly could induce the foraging behavior of roots enabling them to penetrate further into the subsoil for nutrients during the crop’s early growth stage.This would result in a substantial amount of subsoil water being utilized by large,deep roots at later growth stages[20,21,26].Certain cultivars with larger root systems that penetrate the subsoil may obtain more subsoil water in drylands[22,26,27].Therefore,optimizing agronomic practices and cultivar selection could be an effective and practical approach to improve winter wheat yield in the Loess Plateau.

The Weibei dryland is located in the southern part of the Loess Plateau,where the average annual precipitation ranges from 500 to 600 mm,however,only 30%–40% of the rainfall occurs during the growing season of winter wheat[10,28].The present average yield of winter wheat is 4243 kg ha-1,and 36%of smallholder farms could exceed 4920 kg ha-1and even reach 6060 kg ha-1[11].In this area,the widely adopted wheat cultivar is‘‘Changhan 58”(CH58),which was released in 2004.In general,the seeding rate of this cultivar ranges from 120 to 150 kg ha-1and row spacing is 20 cm,and a higher N rate is applied as basal but a lower rate of top-dressing is applied at late growth stages[27,29,30].As a large amount of water is retained in the subsoil at harvest,and soil evaporation accounts for approximately a third of soil water consumption during the growing season in this area[31,32],we hypothesized that grain yield and WUE could be further improved by enhancing the use of subsoil water and decreasing soil evaporation.

The present study investigated the possibility of improving the yield of winter wheat by adjusting agronomic practices and changing cultivars.First,row spacing was decreased from 20 to 10 cm,which was expected to reduce soil evaporation by increasing the area of crop canopy during the early growth stage.Second,the seeding rate was increased from 150 to 180 kg ha-1,which was expected to not only increase the number of deep roots for increased extraction of subsoil water but reduce soil evaporation by increasing plant population.Third,basal N rate was decreased and top-dressing N rate was increased,which was expected to induce root foraging behavior leading to increased penetration of the subsoil where more water would be extracted.Finally,cultivar CH58 was replaced by a new cultivar‘‘Changhang 1”(CH1),which has a larger root system in the subsoil compared with that of CH58 and was expected to absorb more water from subsoil.

2.Materials and methods

2.1.Experimental site

The field experiment was carried out from September 2017 to July 2019 at Changwu Agro-ecological Experiment Station on the Loess Plateau(107° 44.70′E,35° 12.79′N(xiāo)),Changwu,Shaanxi province,China.The experimental station is located at an altitude of 1220 m,with an average annual temperature of 9.1 °C and a frost-free period of 171 days.There is no irrigation in this area because the ground water depth reaches 50–80 m[33].The soil texture is silty loam.The soil properties at 20 cm depth are as follows:pH of 7.8,soil bulk density of 1.3 g cm-3,soil organic matter of 9.98 g kg-1,NO3––N of 6.31 mg kg-1,NH4+–N of 0.99 mg kg-1,available phosphorus of 6.74 mg kg-1,and available potassium of 117.5 mg kg-1.The average annual rainfall of Changwu was 527 mm from 1994 to 2017.During the experiment period,the annual rainfall was 505 mm and 671 mm,from 2017 to 2018 and 2018 to 2019,respectively(Fig.1).According to the drought index(DI)[9],2017–2018 was assessed as a normal year(-0.35＜DI＜0.35)and 2018–2019 was considered a wet year(DI＞0.35).

2.2.Experimental design

We designed five treatments for the experiment which were:1)PF,present farming practices of the high-yielding farmers;2)NR,narrow row spacing:compared with the 20 cm row spacing of PF,the row spacing of NR was 10 cm,and seeds were sowed at the bottom of wider-width ditches;3)HS,high seeding rate:compared with PF,seeding rate was increased by 20%;4)DBN,decreased basal N rate and increased top-dressing N rate:compared with PF(60/40,basal/top-dressing N rate),40% of the total N was applied at sowing time while 60% was top-dressed at jointing stage(40/60,basal/top-dressed N rate);5)NC,new cultivar:compared with PF,the old cultivar‘‘CH58”was replaced by a new cultivar‘‘CH1”.As farmyard manure was widely applied in high-yielding farmland due to its positive effects on soil physical and chemical properties in this area,manure was applied in all treatments.The detailed experimental design of the five treatments is shown in Table 1 and Fig.2.

Table 1The detailed experimental designs of the five treatments.

Table 2The effects of different treatments on aboveground biomass,spikes ha-1,kernels spike-1,1000-grain weight,harvest index,ear-bearing tiller percentage,grain yield and WUE.

The experiment was set up in 2017 in a randomized complete block design with each treatment replicated thrice as plots.The location of the plots was fixed in the two consecutive years.Each plot had an area of 32 m2(4×8 m)and the row spacing was 20 cm(except the row spacing in the NR treatment).Seeds of winter wheat(Triticum aestivumL.)were sown in late September in 2017 and 2018 and harvested in late June of the following year.Split-application of N fertilizer was practiced;the first application was before plowing and top-dressed at the jointing stage.Phosphatic fertilizer and manure were applied once before plowing each year.The manure was sheep dung containing 228 g kg-1organic matter,2.31 mg kg-1NO3-–N,1.23 mg kg-1NH4+–N,and 178.6 mg kg-1available P.

Soil and plant samples were collected at sowing,wintering,regreening,jointing,flowering,and harvesting stages.Regreening,the air temperature gradually rises above 3°C and leaf color changes from gray-green to bright-green(around March 15th to 20th in Changwu).Jointing,the first node of the stem is 1.5–2 cm high above the ground in over 50% of the seedlings in the field(around April 10th to 15th in Changwu).Flowering,the stage is marked by the extrusion of anthers from the spikelet in more than 50%of the tillers in the field(around May 10th to 15th in Changwu).

2.3.Soil water content,soil water storage and evapotranspiration

The oven-drying method was used to determine soil water content(WC)[34].During sowing,regreening,jointing,flowering,and harvesting,soil samples were collected for WC determination at 10 cm intervals(for the 0–1 m soil layer)and 20 cm intervals(for the 1–3 m soil layer).Each treatment included six replicates.Soil water storage(SWS)was calculated according to the following equation[35],where BD is the soil bulk density(g cm-3),which was measured according to Burgess et al.[35];Dis the soil depth(mm).

Fig.1.The monthly rainfall(A)and temperature(B)during 2017–2018 and 2018–2019 seasons.S,sowing time;R,regreening;J,jointing;F,flowering;H,harvesting.

The soil water balance method was used to determine evapotranspiration(ET)as follows[36],

wherePis the effective rainfall,＞5 mm;ΔSWS is the change of soil water storage during a period,mm;Ris the surface runoff(mm),which was left out in this study because the experimental plots had flat surfaces and high ridges.Dis the deep leakage(mm),which was also neglected because almost all of the rainfall that infiltrated into the soil was stored in 0–3 m soil depth in this region[6,30].Therefore,ET was calculated as follows,

2.4.Tiller number and aboveground biomass

Two fixed rows(1 m length per row)were selected and signed at the center of each plot for tiller number counting at the wintering,regreening,jointing,and flowering stages.The aboveground plants were collected from another two rows(1 m length per row)of the same plot and were oven-dried for 48 h at 80 °C for aboveground biomass determination.Each treatment included three replicates.

2.5.Yield,WUE and yield components

Fig.2.Sketch of different planting patterns.PF,present farming practices of the high-yielding farmers;NR,narrow row spacing.

During harvesting in 2018 and 2019,a 3 m2area with two replicates from each plot was harvested for grain yield and yield components determination.Spikes ha-1were determined by counting the number of stems,and 60 spikes were collected from each plot to count kernels spike-1.The grain yield,aboveground biomass,and thousand-grain weight were weighed after the aboveground parts were air-dried.The ear-bearing tiller percentage is equal to the number of spikes divided by the maximum tiller number.WUE was calculated as follows[36],

2.6.Statistical analysis

The SPSS 20.0 software(IBM Company,Chicago,IL,USA)was used to analyze all data.The significance of treatments,year,and treatment×year were analyzed using a mixed ANOVA model.The least significant difference(LSD)atP＜0.05 was used for mean separation.

3.Results

3.1.Yield,WUE and yield components

The grain yields of all treatments in the wet year(2018–2019)were higher than that of the normal year(2017–2018)(Table 2).Compared with PF,the yield of NC and DBN were 10%–15%,and 10%–12% higher,respectively,in the two consecutive years and the yield of HS was 14% higher only in the normal year(2017–2018).WUE increased by 8% in the normal year under NC and was increased by 11% in both years under DBN compared to PF.Aboveground biomass,kernels spike-1,and ear-bearing tiller percentage were influenced by year(Table 2).No significant differences were found in aboveground biomass among the different treatments except the HS treatment which was 12% higher than that of the other treatments in 2017–2018.The number of spikes ha-1of HS was higher than that of the other four treatments.Compared with PF,kernels spike-1of NC was 5%–10% higher in both years,while the kernels spike-1of HS was 6%–12% lower.The 1000-grains weight was not affected by agronomic practices except for a 6%–7% increase under DBN compared with that of PF.In addition,NC and DBN maintained the highest harvest index among the five treatments throughout the two growing seasons.The ear-bearing tiller percentage was higher under DBN during 2017–2019 compared with that under PF;in contrast,NR had the lowest ear-bearing tiller percentage among all treatments(Table 2).

Fig.4.The effects of different treatments on the aboveground biomass at of wintering,regreening,jointing and flowering stages in 2017–2018 and 2018–2019.PF,present farming practices of the high-yielding farmers;NR,narrow row spacing;HS,high seeding rate;DBN,decreased basal N rate and increased top-dressing N rate;NC,new cultivar.The significant differences among five treatments during the same growth stage were denoted by different lowercase letters(P＜0.05).*Indicates there was significant difference between different years in the flowering stage.

3.2.Dynamic changes in tiller number and aboveground biomass

Fig.5.The effects of different treatments on soil water content at different growth stages in 2017–2018 and 2018–2019.PF,present farming practices of the highyielding farmers;NR,narrow row spacing;HS,high seeding rate;DBN,decreased basal N rate and increased top-dressing N rate;NC,new cultivar.Dotted lines is used to make the variation of soil water content at different growth stages more intuitive;small horizontal lines indicate that the soil water content has significant difference among five treatments in the same soil layer(P＜0.05).

The dynamic changes in tiller number among different treatments are shown in Fig.3.The tiller number was higher in the wet year(2018–2019)than in the normal year(2017–2018).Among the five treatments,the highest tiller number throughout the experimental period was observed under the HS treatment,and NR had a higher tiller number compared with PF before the flowering stage.The DBN treatment had the lowest tiller number before the flowering stage among the five treatments for each year but reached levels similar to PF at the flowering stage.Additionally,the aboveground biomass during the wet year was higher at the flowering stage compared to the normal year(Fig.4).The effect of the HS treatment on aboveground biomass was consistent with its effect on tiller number,resulting in 9%–23%higher aboveground biomass during the two years.The aboveground biomass of DBN was significantly decreased by 6%–14% compared with that of PF at the jointing stage in both growing seasons.

Fig.3.The effects of different treatments on the tiller number at wintering,regreening,jointing and flowering stages in 2017–2018 and 2018–2019.PF,present farming practices of the high-yielding farmers;NR,narrow row spacing;HS,high seeding rate;DBN,decreased basal N rate and increased top-dressing N rate;NC,new cultivar.The significant differences among five treatments during the same growth stage were denoted by different lowercase letters(P＜0.05).The capital letters W,R,J and F represent wintering,regreening,jointing and flowering stages,respectively.*Above the capital letters indicates there was significant difference between different years in the same growth stage.

3.3.Dynamic change in soil water content

The soil water content of the different treatments during different growth stages in the 0–3 m soil layer is shown in Fig.5.During the 2017–2018 growing season,slight differences in soil water content were observed among the different treatments at the regreening stage(Fig.5b);soil water content of HS in the 1–2 m soil layer(Fig.5c)was significantly lower at the jointing stage,and also lower in the 1–3 m soil layer at the flowering stage compared to PF(Fig.5c).The soil water content was highest under DBN in the 60–180 cm soil layer among the five treatments(Fig.5d).During harvesting,soil water content under NC,HS,and DBN in the 2–3 m soil layer was significantly lower than that of PF,whereas in the 1–2 m soil layer,it was higher under DBN compared to PF(Fig.5e).During the 2018–2019 growing season,the differences in soil water content among the five treatments remained low at the sowing and regreening stages(Fig.5f,g).The HS treatment had a lower soil water content in the 1–2 m soil layer at the jointing stage and significantly lower soil water content in the 1–3 m soil layer at the flowering stage than that of PF(Fig.5h).In contrast,DBN maintained the highest soil water content in the 0–3 m soil layer among the five treatments(Fig.5i).During harvesting,compared with PF,soil water content was lower in the 2–3 m soil layer under NC,in the 1–3 m soil layer under HS,and in the 2–3 m soil layer under DBN(Fig.5j).

3.4.The effects of different treatments on soil water storage and ET

Soil water storage was significantly higher in the wet year than in the normal year throughout the growing season(Table 3).Compared with PF,soil water storage was significantly lower under HS but higher under DBN at the flowering stage,whereas it was lower under HS and NC during harvesting.The ET was divided into three periods as shown in Table 3,and significant differences were found between years from regreening to harvesting.However,no differences were observed across all treatments in the two years from sowing to regreening.During the 2017–2018 growing season,HS significantly increased ET by 21 mm whereas DBN significantly decreased ET by 20 mm from regreening to flowering compared with PF.From flowering to harvest,NC,HS,and DBN increased ET by 18,13,and 20 mm,respectively,compared with PF.During the 2018–2019 growing season,from regreening to flowering,HS increased ET by 15 mm but DBN reduced ET by 13 mm when compared with PF.In contrast,from flowering to harvest,only DBN significantly increased ET by 24 mm compared with PF.

3.5.Soil water consumption(SWC)in different soil layers during growth stages

Based on the SWC characteristics,we divided SWC into three periods,and investigated the changes in SWC under three soil layers(0–1 m,1–2 m,and 2–3 m)(Fig.6).During the 2017–2018 growing season,from sowing to regreening,the differences in SWC among the five treatments were small,and the soil water storage in the 1–3 m soil layer was supplemented in all treatments(Fig.6A a–e).From regreening to flowering,HS increased SWC by 7 mm in the 1–2 m soil layer and by 11 mm in the 2–3 m soil layer compared with PF,whereas DBN decreased SWC by 13 mm in the 0–1 m soil layer and by 9 mm in the 1–2 m soil layer(Fig.6A f–j).From flowering to harvest,NC,HS,and DBN significantly increased SWC by 13 mm,16 mm,and 11 mm,respectively,in the 2–3 m soil layer compared with PF(Fig.6A k–o).During the 2018–2019 growing season,from sowing to regreening,the SWC among the five treatments were similar(Fig.6B a–e);from regreening to flowering,in comparison with PF,the SWC under HS in the 1–2 and 2–3 m soil layers was 13 mm and 8 mm higher,respectively,but 10 mm lower in the 0–1 m soil layer(Fig.6B f–j).From flowering to harvest,NC,HS,and DBN increased SWC by 8 mm,9 mm and 13 mm,respectively,in the 2–3 m soil layer compared with PF(Fig.6B k–o).

4.Discussion

Our two-year study showed that the present grain yield of rainfed winter wheat in the Loess Plateau could be further improved by optimizing agronomic practices or changing cultivar.Generally,enhancing WUE is a primary principle used to increase crop yield in rainfed agriculture[10].Compared with PF,WUE under DBN and NC were enhanced,which was due to the higher HI and the different ET patterns.The different ET patterns were attributed to the different SWCs,especially subsoil water consumption after flowering.Under DBN,the total ET was not increased,but water consumption was decreased in the 0–2 m soil layer before flowering and was increased in the 2–3 m soil layer after flowering.The total ET of NC was enhanced because it consumed more subsoil water(2–3 m)after flowering.In contrast,NR could not enhance WUE and yield due to a failure to regulate soil water consumption.HS cannot increase yield steadily every year because there is the risk of reducing kernels spike-1due to the high plant population.In addition,the grain yield and WUE were significantly higher in the wet year than in the normal year due to the higher aboveground biomass accumulation and the greater ET after flowering.

Under NR,seeds were sowed in ditches with a wider-width bottom at the expense of narrower rowing spacing,which provided broader space for seeds to grow eventually leading to increased crop stands per unit area.Subsequently,the increased crop stands per unit area would accelerate the early achievement of the critical leaf area index(LAI)and thus result in increased competition for nutrients and water among individual plants[37–40].Therefore,some studies were successful in improving yield by increasing the population and nutrient use efficiency under NR[23,41].In contrast,theoretically,a high LAI during the early growth stage will contribute to reduced soil evaporation and simultaneously increase transpiration[19,23].Compared with PF,ET did not change significantly under NR during the early growth stage,indicating that the decreased soil evaporation and increased transpiration caused by NR were equivalent.More water available for transpiration translates to higher biomass production or plant population and,therefore,greater potential yield[23].A higher tiller number was observed under NR at the regreening stage,but it gradually decreased to levels similar to those of PF at a later stage,suggesting that the increased population under NR in the early stage did not convert into yield in the later stages.Ultimately,the SWC,ET,WUE,and yield of NR did not change compared with those obtained under PF(Tables 2,3;Fig.6).These results imply that reducing soil evaporation by NR during the early growth stage could not improve WUE and yield effectively in this area.

A relatively high seeding rate may lead to high spike numbers and high dry matter accumulation,which contribute to achieving high yields in wheat production[42–44].Increasing the seeding rate could also enlarge root length density in the subsoil,consequently absorbing more water[22,25,45].As expected,HS resulted in higher plant populations and increased consumption of subsoil water over the entire growing season(Figs.3,6).Thus,WUE was not increased under HS compared with PF due to the higher ETboth before and after flowering.The grain yield of HS improved in the 2017–2018 season,but not in 2018–2019 due to the acute decline in kernels spike-1,in comparison with that of PF(Table 2).Large plant populations can lead to a decrease in kernels spike-1in wheat production,which is always a risk to be considered under high seeding rates[42].In contrast,in dryland agriculture,the excessive water consumption of large plant populations during early growth can be exacerbated by low initial soil water storage and insufficient rainfall in the growing season.Together,these factors increase the risk of severe water deficit in later growth stages[30,46].Therefore,although increasing the seeding rate could consume more subsoil water and increase plant population,it cannot enhance yield steadily every year.

Table 3The effects of different treatments on soil water storage and ET.

Fig.6.Effects of different treatments on soil water storage and soil water consumption among sowing–regreening,regreening–flowering and flowering–harvest during 2017–2018(A)and 2018–2019(B).PF,present farming practices of the high-yielding farmers;NR,narrow row spacing;HS,high seeding rate;DBN,decreased basal N rate and increased top-dressing N rate;NC,new cultivar.Dotted and solid lines represent the soil water content(w/w,%)of different treatments at sowing time,regreening stage,and flowering stage or harvest time.The soil was divided into three layers,0–1 m,1–2 m and 2–3 m.The residual soil water storage(mm)of three layers at the end of each growth period is shown by cyan(0–1 m)and light green(1–3 m)areas;the change of soil water storage,namely,soil water consumption(mm)in three soil layers were represented by blue areas(supplemented soil water storage)or peach areas(consumed soil water storage).Numbers in PF(A,a,f,k;B,a,g,m)presented on the left side of the solid line represent the residual soil water storage at the end of each growth period,on the right side of the solid line represent the change in soil water storage,either an decrease(red numbers)or increase(blue numbers).Numbers in other treatments presented on the right side of the solid line represent the increases(positive value)or the decreases(negative value)of soil water storage under other treatments relative to PF in the same soil layer during the same growth period.*Significant at P＜0.05.

The development of plant populations depends on soil nutrients and water supply[35].In this study,the soil nutrient supply of DBN was low during the early growth stage but increased by the late growth stage.Although low soil nutrients limited the development of plant populations in the early stages,the tiller number of DBN after flowering showed no differences due to the higher earbearing tiller percentage compared with PF(Table 2;Fig.3).The lower plant population under DBN during the early growth stage also decreased the consumption of soil water before flowering(10–22 mm lower in the 0–2 m soil layer).As a result,the soil water content during flowering was higher,implying that there was sufficient water available for wheat growth(11–13 mm higher in the 2–3 m subsoil)under DBN than under PF during the late growth stage(Figs.5,6).Although the total ET under DBN over the whole growing season was not increased,better water supply for wheat growth after flowering led to higher WUE and yield of the DBN treatment compared to those of PF.Generally,decreasing the basal N rate and increasing the top-dressing N rate could be a useful agronomic practice for increasing winter wheat yield in this area.It should be noted that,if the initial soil nutrients are low,a low basal N fertilizer rate may drastically limit the development of the plant population during the early growth stage.As a consequence,even if the soil nutrients and water supply are sufficient during the late growth stage,the wheat grain yield would not increase but would instead decrease due to a smaller plant population[47–49].Therefore,applying too low amounts of basal N fertilizer has the potential risk of reducing yields in poor soils.

A well-developed,deep root system might extract more subsoil water for crop growth in drylands[26].Moreover,the use of soil water,especially subsoil water during the late growth stage is of great value to grain yield formation and increases the WUE of wheat in dryland farming[15,16,50].In the present study,compared with CH58(PF),the new cultivar CH1(NC)consumed more subsoil water(8–13 mm in the 2–3 m soil layer)after flowering(Fig.6),indicating a better water supply for growth of CH1 after flowering.Our results are in agreement with a previous study in which the CH1 cultivar was reported to have both high root biomass and root length density in the subsoil compared with the CH58 cultivar,and possessed a higher yield potential in this area[22].

Crop yield and WUE were significantly affected by the harvest index,which was not only related to cultivar characteristics,but also could be regulated by agronomic practices[51].Previous studies have also reported that the amount of ET after flowering may influence the harvest index of wheat[10,52].In the current study,the HI and ET in the DBN treatment after flowering were higher than those of PF,indicating that DBN was beneficial in improving HI.NC also had higher HI and ET after flowering,which suggested that the ET after flowering should be considered when selecting cultivars for use in dryland agriculture(Tables 2,3).

Only the soil water within the 3 m soil layer was considered in this study.In fact,in the Loess Plateau,the soil water is moving upward during most of the year under the effect of high vapor pressure deficit(VPD),and it has been shown that subsoil water(2–3 m)would likewise be affected by VPD and move upwards due to the strong water conductivity of the silty loamy soil[32].Therefore,there is a possibility that even the water below the 3 m soil layer could be used by wheat.In this study,the net biomass production from flowering to harvest was approximately 992–1363 kg ha-1in 2017–2018 and 400–1010 kg ha-1in 2018–2019.Net photosynthesis after flowering contributed significantly to wheat yield,although the proportion was much lower in the dryland.However,compared with most studies,the net biomass production after flowering was much lower in this study,which may be due to more cloudy days after flowering in the two experimental years.It also implies that the yield of this area could be higher in years with a good climate.

5.Conclusions

Our study demonstrated that in the Loess Plateau,optimizing agronomic practices could further increase the present rainfed winter wheat yield.DBN and NC significantly increased the yield for the two experimental years compared with PF.The increase in yield was due to the greater use of subsoil water and the improvement of WUE.Under DBN,soil water consumption was reduced before flowering but increased after flowering,especially in the subsoil(2–3 m).Cultivars with more deep roots(NC)may absorb more subsoil water(2–3 m)after flowering.Taken together,the results of this study suggest that winter wheat yield in the Loess Plateau could be further increased by improving the use of soil water,especially the use of subsoil water after flowering.Moreover,in this study,only a single agronomic practice was considered for each comparison with PF;the effect of integrated agronomic practices(seeding rate,N application)using different cultivars on wheat yield are worth investigating in the future.

CRediT authorship contribution statement

Shiwen Wangconceived and designed the study.Wenjia Yang,Weijian Liu,and Yulin Licarried the experiments and collected the data.Wenjia Yanganalyzed the data and drafted the manuscript.Lina Yin and Xiping Dengprovided suggestions and polished the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the National Basic Research Program of China(2015CB150402),the National Key Technology R&D Program(2015BAD22B01),and 111 project of Chinese Education Ministry(B12007).