XlAO Jing-xiu,ZHU Ying-an,BAl Wen-lian,LlU Zhen-yang,TANG Li,ZHENG Yi,

1 College of Resources and Environment,Yunnan Agricultural University,Kunming 650201,P.R.China

2 College of Horticulture and Landscape,Yunnan Agricultural University,Kunming 650201,P.R.China

3 Yunnan Open University,Kunming 650223,P.R.China

Abstract Yield performance in cereal and legume intercropping is related to nutrient management,however,the yield response of companion crops to nitrogen (N) input is inconclusive and only limited efforts have focused on rationed phosphorous (P)fertilization.In this study,two multi-year field experiments were implemented from 2014-2019 under identical conditions.Two factors in a randomized complete block design were adopted in both experiments.In field experiment 1,the two factors included three planting patterns (mono-cropped wheat (MW),mono-cropped faba bean (MF),and wheat and faba bean intercropping (W//F)) and four N application rates (N0,0 kg N ha-1;N1,90 and 45 kg N ha-1 for wheat and faba beans,respectively;N2,180 and 90 kg N ha-1 for wheat and faba beans,respectively;and N3,270 and 135 kg N ha-1 for wheat and faba beans,respectively).In field experiment 2,the two factors included three P application rates (P0,0 kg P2O5 ha-1;P1,45 kg P2O5 ha-1;and P2,90 kg P2O5 ha-1) and the same three planting patterns (MW,MF,and W//F).The yield performances of inter-and mono-cropped wheat and faba beans under different N and P application rates were analyzed and the optimal N and P rates for intercropped wheat (IW) and MW were estimated.The results revealed that intercropping favored wheat yield and was adverse to faba bean yield.Wheat yield increased by 18-26%,but faba bean yield decreased by 5-21% in W//F compared to MW and MF,respectively.The stimulated IW yield drove the yield advantage in W//F with an average land equivalent ratio (LER) of 1.12.N and P fertilization benefited IW yield,but reduced intercropped faba bean (IF) yield.Nevertheless,the partial LER of wheat (pLERwheat) decreased with increasing N application rates,and the partial LER of faba bean (pLERfababean) decreased with increasing P application rates.Thus,LER decreased as N input increased and tended to decline as P rates increased.IW maintained a similar yield as MW,even under reduced 40-50% N fertilizer and 30-40%P fertilizer conditions.The estimated optimum N application rates for IW and MW were 150 and 168 kg ha-1,respectively,and 63 and 62 kg ha-1 for P2O5,respectively.In conclusion,W//F exhibited yield advantages due to stimulated IW yield,but the intercropping yield benefit decreased as N and P inputs increased.Thus,it was concluded that modulated N and P rates could maximize the economic and ecological functions of intercropping.Based on the results,rates of 150 kg Nha-1 and 60 kg P2O5 ha-1 are recommended for IW production in southwestern China and places with similar conditions.

Keywords:land equivalent ratio,nitrogen and phosphorus,optimal application rate,wheat and faba bean intercropping,yield performance

1.lntroduction

Legume and cereal intercropping is a traditional agricultural practice in China that plays an important role in modern Chinese agriculture due to the yield advantages of intercropping (Knörzeret al.2009;Machado 2009;Honget al.2017).Over the past few decades,numerous studies have demonstrated the benefits of legume and cereal intercropping,including the reduced risk of crop failure,enhanced income for smallholder farmers,and food and nutrition security in vulnerable production systems(Liet al.2009;Ngwiraet al.2012a;Rusinamhodziet al.2012).In developing countries,farmers have widely adopted intercropping practices because both improved and stabilized crop yields are obtained when cereals are intercropped with legumes (Hauggaard-Nielsenet al.2009;Liet al.2009).Recently,legume and cereal intercropping has been increasingly promoted due to its benefits of improved agricultural sustainability,sufficient food and feed production levels,and sustained land productivity and resilience (Bedoussacet al.2010;Martin-Guayet al.2018);additionally,it provides a pathway for ecological intensification in agriculture (Cuiet al.2019;Funget al.2019;Raji and Dörsch 2019).

Numerous studies have been conducted to better understand the underlying mechanisms of intercropping,and have reported that the advantages of intercropping are due one or more of the following factors:above-and belowground interactions,niche complementarity,species-species interactions,disease and pest avoidance mechanisms,and microtopographic features (Liet al.2007,2014;Bedoussacet al.2010;Heet al.2013;Brookeret al.2015;Qiaoet al.2016).A meta-analysis quantified intercropping yield advantages;however,conflicting results were obtained when the land equivalent ratio (LER) was used to represent overyielding in intercropping (Pelzeret al.2014;Bedoussacet al.2015;Yuet al.2016;Martin-Guayet al.2018).Previously,it was reported that most combinations of cereal and legume crops (＞90%) exhibited higher grain yields than sole crops(LER＞＞1) (Pelzeet al.2014).However,intercropping with higher LER values does not always equate to higher yield performances,as the grain yields of companion crops differ under different intercropping combinations (Agegnehuet al.2008;Bedoussacet al.2015).Moreover,the LER index has rarely been explored for demonstrating the intercropping yield component (Bedoussacet al.2011).It is possible that the LER value in combination with the original grain yield could be used for evaluating intercropping yield performance,which is imperative for developing guidelines regarding specific intercropping combinations.

Legume-based intercropping takes advantage of the nitrogen (N)-fixing ability of legumes to improve the nutrient use efficiency and reduce fertilizer application (Liet al.2003;Chenet al.2015;Zhouet al.2017;Xiaoet al.2018).When wheat was intercropped with winter red clover,reduced N application rates (10-12%) had no effect on wheat grain yield(Gaudinet al.2014).Moreover,25-50% N fertilizer was retained when sorghum and pearl millet were intercropped with soybean (Layeket al.2014).Similarly,the N fertilizer need was reduced by 5-15% when wheat was intercropped with faba bean because approximately 5% N was fixed by faba bean and transferred to wheat (Xiaoet al.2004,2018).In addition to N,legume and cereal intercropping also benefits crop phosphorus (P) uptake and soil P mobilization(Liet al.2007;Wanget al.2017).Moreover,interspecies N competition in legume-based intercropping systems was found to be regulated by P rates,as P fertilizer application stimulated legume nodulation formation (Isaacet al.2012).Clearly,nutrient (N and P) management in intercropping systems differs from corresponding mono-cropped systems.

Wheat and faba bean intercropping is a good agricultural practice in Yunnan Province,located in southwestern China,as it has been reported to increase crop yield,improve disease resistance,and promote crop nutrient uptake and utilization (Xiaoet al.2004;Chenet al.2007;Liet al.2014).However,limited attention has been paid to the yield component when wheat is intercropped with faba beans under long-term field conditions;thus,the roles of N and P in the increasing yield resulting from intercropping remain unclear.A previous study found that wheat and faba bean intercropping could reduce the amount of N fertilizer needed due to the biological N fixation ability of faba beans (Xiaoet al.2018).However,N and P management in intercropping is conducted in accordance with corresponding monocropped practices.Thus,the objectives of this study were to:1) quantify the yield performance of a wheat and faba bean intercropping system and evaluate the effects of N and P application rates on intercropping yield performance based on a five-year field experiment,and 2) identify the optimal N and P application rates of intercropped wheat and faba beans to enhance intercropping ecological intensification.

2.Materials and methods

2.1 Experiment site

The field experiments were conducted at the Yunnan Agricultural University Research Station from 2014-2019.The station is located in Xundian (23°32´N,103°13´E) at an altitude of 1 953 m and is 75 km northeast of Kunming City,Yunnan Province,China.The climate is categorized as subtropical monsoon with an average temperature of 14°C and annual precipitation of 1 040 mm.The average monthly temperatures and monthly precipitation amounts during the experiment are shown in Fig.1.The soil was cultivated red soil,which originated from sandstone and shale.Corn had been planted for several years before the experiment was established.The two field experiments were carried out in identical fields.At the beginning of the present study,the field soil bulk density was 1.36 g cm-3and the pH was 7.2.The soil organic matter and total N,P,and potassium (K) contents were 35.1,1.11,0.73,and 17.68 g kg-1,respectively.The available N (NaOH hydrolyzed),K(NH4OAc-exchangeable),and P (Olsen-P) values were 80,146,and 17 mg kg-1,respectively.

2.2.Experimental design

Field experiment 1Four N application rates (no N fertilizer(N0),90 kg N ha-1(N1),180 kg N ha-1(N2),and 270 kg N ha-1(N3) for wheat;no N fertilizer (N0),45 kg N ha-1(N1),90 kg N ha-1(N2),and 135 kg N ha-1(N3) for faba bean) and three planting patterns (mono-cropped wheat(MW),mono-cropped faba bean (MF),and wheat and faba bean intercropping (W//F)) were arranged in a randomized complete block design with three replications.The N application rates of mono-and inter-cropped wheat and faba bean were identical in each plot.

In field experiment 1,a total of 12 treatments and 36 plots were established in 2014.The plot size was 5.4 m×6 m(32.4 m2).The inter-row distance of wheat was 0.2 m with 27 rows in each mono-cropped plot,and the seed rate was 180 kg ha-1.For faba beans,the inter-row distance was 0.3 m and the interplant space was 0.1 m;the plant density was 3.3×104plants ha-1with 18 rows of faba beans in each mono-cropped plot.

The planting pattern of W//F was in accordance with field practices,i.e.,six rows of wheat were intercropped with two rows of faba beans.We adopted the same plant density for mono-and inter-cropped wheat and faba bean in this study,and the row spacing between wheat and faba bean in the resulting intercropping plot was 25 cm.Thus,there were three strips of W//F in each plot that contained 18 rows of wheat and six rows of faba beans.Wheat and faba bean occupied 2/3 and 1/3 of the total land area in each intercropping plot,respectively.

Urea was used as the N fertilizer in field experiment 1.For mono-and inter-cropped wheat,the N fertilizer was divided into two applications,i.e.,half of the total amount for each treatment (0,45,90,and 135 kg N ha-1,respectively)was applied as basal fertilizer,and the second half of the N rate was applied as topdressing at the wheat jointing stage.For mono-and inter-cropped faba bean,all N fertilizer for each treatment (0,45,90,and 135 kg N ha-1,respectively)was totally applied as broadcast.Superphosphate and potassium sulphate were used as the P and K fertilizers,respectively.The application rates were 90 kg P2O5ha-1and 90 kg K2O ha-1for each crop,and they were applied as base fertilizers according to local farming practices.Equal portions of fertilizer were partitioned to each row in a given plot for each treatment,and all fertilizers were weighed for each row separately.For a given intercropping plot,the fertilizer application rate was identical to the corresponding mono-cropped plot based on the same plot area for both wheat and faba bean,and all fertilizers were evenly applied to each row by hand.The second half of the N fertilizer as wheat topdressing was only applied to wheat rows by hand at the wheat jointing stage in a given intercropping plot.Sowing and fertilization throughout the experiments were conducted by hand.

Field experiment 2In field experiment 2,a randomized complete block design involving two factors was also used and conducted in triplicate.The first factor included three P application rates (no P fertilizer (P0),45 kg P2O5ha-1(P1),and 90 kg P2O5ha-1(P2)) and the second factor consisted of the three planting patterns (MW,MF,and W//F).

In total,18 treatments were applied in field experiment 2 beginning in 2014.The plot size,plant density,intercropping pattern,varieties of crops,fertilizer chemical forms,and other crop management practices were in accordance with field experiment 1.P fertilizer was applied as basal fertilizer before sowing.Additionally,90 kg N ha-1and 90 kg K2O ha-1were applied as broadcast in all of the treatments.Another 90 kg N ha-1was applied to MW and IW at the jointing stage as topdressing.

2.3.Crop management and sample collection

The varieties of wheat (TriticumaestivumL.) and faba bean (ViciafabaL.) were Yunmai 52 and Yuxi Dalidou,respectively.Seeds were provided by the Yunnan Academy of Agricultural Sciences.Wheat and faba beans were sown from the 20th-30th of October during 2014-2018 and harvested from the 10th-20th of April during the next year after sowing.After wheat and faba beans were harvested,each plot was kept fallow from May to September.Irrigation and plant protection management was in accordance with local farming practices.Crop management was uniform across all treatments.In each year,the grain yields of wheat and faba beans in each plot were measured.

2.4.Data collection

LERLER is dependent on the yield of a cereal/legume intercrop system and is equivalent to the sum of the partial LER values for the cereal and legume crops (Willey 1990;Bedoussacet al.2015).LER was calculated using the following equations:

where pLERwheatand pLERfababeanrepresent the partial LER values of wheat and faba beans,respectively,Yintwand Yintbare the component yields of wheat and faba beans in a given intercropping plot,and Ymonowand Ymonobare the yields of MW and MF in a given mono-cropped plot.When LER＞1,intercropping has a yield advantage;and when LER＜1,intercropping has no yield advantage (Willey 1990).

Yield ratioYield ratio (Ratioyield) was used to compare the grain yield of intercropping per unit area with the grain yield of MW and MF per unit area (Pelzeret al.2014).Ratioyieldwas calculated using the following equations:

where Yintw,Yintb,Ymonow,and Ymonobare defined in eqs.(2) and (3),LAintwand LAintbare the land areas of IW and IF,respectively,and LAmonowand LAmonobare the land areas of MW and MF,respectively.In the present study,Pw is 2/3 and Pb is 1/3.

Linear-plateau modelThe linear-plateau model was fit to the data for the five years of the field experiment using the NLIN procedure.To eliminate the variability among years,the relative yield (Ry) was calculated using the following equation:

where Yfis the yield of each treatment (kg ha-1) and Ymis the maximum yield (kg ha-1) of each year under mono-and inter-cropping conditions.

The linear-plateau model was defined using the following equations:

whereYis the relative grain yield (kg ha-1),Xis the N or P application rate (kg ha-1),ais the intercept,bis the linear coefficient,Cis the intersection of the straight line and platform,andMis the relative platform yield (%) (Cerrato and Blackmer 1990).

2.5.Statistical analysis

A two-way analysis of variance (ANOVA) was performed using the MIXED procedure with SPSS Software 19.0 (SPSS Inc.,Chicago,IL,USA) to test for significant differences among treatments.Planting patterns and N and P application rates were considered the fixed factors,while replication was considered the random factor.Significant differences among treatments for each trait were investigated using Duncan’s multiple range post hoc test when theF-value was significant (P＜0.05).The variances of LER,pLER,and Ratioyieldamong different N and P application rates were investigated by a one-way ANOVA.The independent samplet-test was used to compare the total intercropping yield and mean yields of MW and MF.Regression analyses were performed for each year to evaluate the effects of N and P application rates on grain yield production,LER,and Ratioyield.Linear,quadratic,and linear-plateau models were used to simulate the regressions of relative grain yield with N and P application rates.The linear-plateau model was the best fit model and explained the IW and MW yield responses of N and P application rates.Thus,the optimal N and P rates were estimated based on the corresponding linear-plateau model with 95 and 100% confidence intervals.

3.Results

3.1.Wheat and faba bean yields in intercropping

Wheat yield was regulated by planting patterns,N and P application rates,and the interaction between N or P application rates and planting patterns across years,except for 2015 in field experiment 2 (Table 1).Faba bean yield was regulated by planting patterns,and N and P application rates,but was seldom affected by the interaction between N or P application rates and planting patterns in the two experiments (Table 1).

In field experiment 1,wheat yield increased by 18-26% and faba bean yield decreased by 5-21% (Fig.2) for wheat intercropped with faba bean as compared to the corresponding mono-cropped yields,regardless of N application rate.Under N0,N1,N2,and N3 levels,IW yields were higher than MW by 38,27,22,and 14%(Fig.2-A),respectively,but IF yields were lower than MF by 9,11,19,and 17%(Fig.2-B),respectively.IW yields under N1 and N2 were significantly higher (P＜0.05)than MW under N2 and N3 due to the interaction between planting patterns and N application rates (Fig.2-A).

Fig.2 Mono-and inter-cropped wheat (A) and faba bean (B) grain yields under different N application rates.MW,mono-cropped wheat;IW,intercropped wheat;MF,mono-cropped faba bean;IF,inter-cropped faba bean.IW and IF yields are the yields of the whole intercropping area.N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.Error bar is SE (n=3).＊,significant difference between mono-and inter-cropped treatments under the same N rates in the same year (P＜0.05).

In field experiment 2,W//F increased wheat yield by 18-26% (Fig.3-A) and decreased faba bean yield by 4-28% (Fig.3-B) when compared to MW and MF,regardless of P application rate.Wheat yields increased by 19,22,and 22%(Fig.3-A),but faba bean yields decreased by 6,14,and 18% (Fig.3-B) for wheat intercropped with faba bean under P0,P1,and P2 levels,respectively,as compared to the corresponding MW and MF.The yield of IW at the P1 level was similar or higher (P＜0.05) than that of MW at the P2 level (Fig.3-A).

Fig.3 Mono-and inter-cropped wheat (A) and faba bean (B) grain yields under different P application rates.MW,mono-cropped wheat;IW,inter-cropped wheat;MF,mono-cropped faba bean;IF,intercropped faba bean.IW and IF yield are the yields of the whole intercropping area.P0,P1,and P2 are P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.Error bar is SE (n=3).＊,difference between mono-and inter-cropped treatments under the same P rates in the same year (P＜0.05).

Fluctuations in the yields of both for mono-and inter-cropped wheat and faba bean were found during 2014-2019 (Figs.2 and 3),and both wheat and faba bean yields tended to decline during the years of the study in both field experiments.Intercropping showed less yield reduction than mono-cropping under the N0 level,but no difference in yield loss between the mono-and inter-cropping systems was found with the application of N and P fertilizer over time.

3.2.Yield components in intercropping

The total W//F grain yield (wheat+faba bean) was significantly higher (P＜0.05) than the mean yield of MW and MF (Fig.4-A and C) based on the same land area in field experiment 1.However,no differences were detected between the total W//F grain yield and MW yield (Fig.4-B).In field experiment 2,the total W//F grain yields were significantly higher (P＜0.05) than MF yields under the different P application rates (Fig.4-E).Compared to the mean yields of MW and MF and the individual yield of MW,total W//F grain yield did not exhibit any yield advantage (Fig.4-E and F).

Fig.4 The total wheat and faba bean inter-cropping grain yields compared to mono-cropped yields under different N (A,B and C)and P (D,E and F) application rates.A and D,function of the mean mono-cropped grain yields under different N (y=1.13x+0.028,r2=0.898) and P rates (y=0.73x+0.66,r2=0.946).B and E,function of mono-cropped wheat yields under different N (y=0.78x+1.06,r2=0.823) and P rates (y=1.16x-0.16,r2=0.879).C and F,function of mono-cropped faba bean grain yields under different N(y=0.78x+1.14,r2=0.416) and P rates (y=0.93x+0.69,r2=0.483).N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.P0,P1,and P2 are P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.

Ratiowheatincreased due to N-fertilization,but no differences were detected among the N1,N2,or N3 levels (Fig.5-A).In contrast,Ratiofababeandecreased as N application increased (Fig.5-B).In field experiment 2,Ratiowheatincreased but Ratiofababeandecreased as P application rate increased (Fig.5-C and D).

Fig.5 Intercropping yields under different N (A and B) and P (C and D) application rates.N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.P0,P1,and P2 are P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.Values with the same letter in each figure panel are not significantly different among N and P application rates (one-way ANOVA,P＜0.05).

3.3.LER

The average pLERwheatwas 0.82 and it ranged from 0.6-1.19 in both field experiments,where 96% of the observations were significantly higher (P＜0.001) than 0.67.pLERwheatincreased due to the N and P application rates (Fig.6-A and D).In contrast,the average pLERfababeanwas 0.29 and it ranged from 0.14-0.45,where 85% of the observations were significantly lower (P＜0.001) than 0.33.Moreover,pLERfababeandecreased as the N and P application rates increased (Fig.6-B and E).

In most cases,the results revealed that pLERwheatwas higher than the wheat land proportion in intercropping(2/3),while pLERfababeanwas lower than the faba bean land proportion in intercropping (1/3) (Fig.7-A and B).Moreover,87% of the experimental units were above the green line(Fig.7-A and B),which corresponds to pLERwheat+pLERfababean=1.Thus,the grain yield production of intercrops was clearly more efficient than the comparable mono-crops.

Fig.7 pLERwheat relationships to pLERfababean under different N (A) and P (B) application rates.LER,land equivalent ratio;pLERwheat and pLERfababean,partial LER values of wheat and faba beans,respectively.The dark line represents pLERfababean=1/3,the green line represents pLERwheat=2/3,the blue line represents pLERwheat=2pLERfababean,and the red line represents LER=1.N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.P0,P1,and P2 are P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.

In total,the results of both field experiments revealed that the total LER was significantly greater (P＜0.05) than 1,with an average of 1.12.The average LER was 1.23±0.12 under N0 and 1.09±0.1 under N1,N2,and N3.LER was not regulated by P application rates,but decreased due to the N rates (Fig.6-C and F).

Fig.6 Land equivalent ratio (LER) under different N (A,B and C) and P (D,E and F) application rates.N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.P0,P1,and P2 are P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.Values with the same letter in each figure panel are not significantly different among the N and P application rates (one-way ANOVA,P＜0.05).

3.4.Optimal N and P fertilizer estimation

The linear-plateau model estimated the optimal N and P application rates for MW and IW based on both field experiments (R2=0.743-0.822,P＜0.001).The optimal N application rates were 168 and 150 kg N ha-1for MW and IW,respectively (Fig.8-A and B).Wheat achieved the same yields as MW per unit area compared to W//F,despite the reduced (40-50%) N application rates.Moreover,the N application rates at 122 kg ha-1reached 95% of the optimal grain yield for MW based on the model.For IW,the value was 110 kg ha-1.

Based on the model,the optimal P application rates were 62 and 63 kg P2O5ha-1for MW and IW,respectively (Fig.8-C and D).Compared to MW,W//F decreased by 30-40% P input,but still achieved the same wheat yield under the same unit area (Fig.8-C and D).According to the model,when the application rates were 49 and 51 kg P2O5ha-1for MW and IW,respectively,MW and IW achieved 95% of the optimal grain yield.

Fig.8 The estimated optimal N (A and B) and P (C and D) application rates for mono-and intercropped wheat based on the linear-plateau model.MW,mono-cropped wheat;IW,inter-cropped wheat.N0,N1,N2,and N3 are N application rates at 0,90,180,and 270 kg ha-1 for wheat and 0,45,90 and 135 kg ha-1 for faba bean,respectively.P0,P1,and P2 were P application rates at 0,45,and 90 kg P2O5 ha-1,respectively.

4.Discussion

4.1.Yield performance in intercropping

According to previous studies,the effects of cereal and legume intercropping on companion crop yields are conflicting.Renet al.(2016) reported that legume and cereal intercropping significantly increased cereal crop yield,while other studies argued that intercropping benefited the yields of both crops (Zuoet al.2004;Labergeet al.2011).In this study,W//F favored wheat yield production,but was adverse to faba bean yield (Figs.2 and 3).Furthermore,the total W//F yield was always higher than the MF yield and frequently higher than the mean yield of MF and MW (Fig.4).These results confirmed that W//F exhibited the higher productivity,which matched the findings of previous studies(Agegnehuet al.2008;Stefaniset al.2016).Moreover,the higher grain yield of W//F was mainly attributed to a stimulated IW yield (Figs.2 and 3).

On average,LER values in cereal and legume intercropping are greater than 1 (Bedoussac and Justes 2011;Pelzeret al.2014;Yuet al.2015;Martin-Guayet al.2018).Generally,legume-based intercropping is more important in low-input and low-yield farming systems than high-input or high-yield cropping systems (Ofori and Stern 1987;Bedoussacet al.2015;Stagnariet al.2017),as the yields that benefit from intercropping decrease due to fertilizer application (Liet al.2007;Yuet al.2015).In this study,the average LER value was 1.12 and it declined as N application rates increased (Fig.6-C).High N availability tends to reduce legume yield,but is beneficial to cereal yield (Hauggaard-Nielsenet al.2009;Naudinet al.2010).In this study,IF yields were greatly reduced compared to the increased IW yields under N2 and N3,which were not beneficial to the intercropping yield advantages (Figs.2 and 3).Clearly,the intercropping yield advantages diminished under N3 (Fig.6-C).The results of a meta-analysis showed that the LER of legume-based intercropping was not affected by N application rates (Pelzeret al.2014;Xuet al.2020).However,we found that the LER value was dependent on N input (Fig.6-C),which was in accordance with the stress gradient hypothesis.In contrast,the LER value was not regulated by P application rate (Fig.6-D),which was similar to that in the meta-analysis of Liet al.(2020);the reason being the high soil P availability (Olsen-P,17 mg kg-1).

The yield responses of wheat and faba beans to N and P application rates were different in this study.Specifically,pLERwheatdecreased as N application rates increased,but pLERfababeanwas not affected by N input(Fig.6-A and B).In contrast,pLERfababeandecreased as P application rates increased,but pLERwheatwas not affected by P input (Fig.6-D and E).Moreover,wheat was the dominant component crop and comprised 2/3 of the land proportion in the intercropping system.Thus,pLERwheatdecreased as N application rates increased,resulting in the decrease of the total LER value,while pLERfababeanhad little effect on total LER as P application rates increased.Notably,IW yield decreases as IF seed rates increased(Agegnehuet al.2008),and crop yield performances change as the spatial structure of intercrops differs (Mohta and De 1980).Thus,the effects of N and P application rates on IW and IF yields and LER values may be different if the proportions of companion crops or strip structures in the intercropping conditions change.

The present study was partially consistent with the observation of Rusinamhodziet al.(2012) in that legumebased intercropping could reduce the risk of crop failure and increase food security in a vulnerable production system,but the positive intercropping effect was altered by N or P application,because we found that intercropping only had less yield loss with years than mono-cropping under the N0 level (Fig.2).Thus,legume-based intercropping is often undertaken by farmers who use under low-input conditions(Ngwiraet al.2012b).

4.2.Companion crop interactions in intercropping

Two companion crops that are complementary in using mineral soil N and atmospheric N2explain the advantages of cereal and legume intercropping (Jensenet al.1996;Hauggaard-Nielsen and Jensen 2001;Hauggaard-Nielsenet al.2009;Bedoussacet al.2010).Typically,cereals are more productive than legumes (Jensen 1996),hence pLERwheatwas frequently higher than 2pLERfababeanin this study (Fig.7).However,wheat and faba bean yields were similar under N0 and P0,while wheat yield was considerably higher than faba beans under increased N and P application rates (Figs.2 and 3).This result was likely due to legumes having higher interspecific competition under lower soil N levels,while the companion cereals competed better under higher soil N levels (Hauggaard-Nielsen and Jensen 2001).These results confirmed that the application of N and P fertilizers promoted the competitive ability of wheat against faba beans for space,nutrients,and solar radiation in intercropping conditions (Layeket al.2014).Moreover,Ratiowheatincreased and Ratiofababeandecreased as N and P inputs increased (Fig.5),indicating a negative association between wheat and faba bean yields in intercropping,which was likely due to a trade-off within the yields of the two crops(Agegnehuet al.2008).

According to the stress-gradient hypothesis (Maestreet al.2009),facilitation is common under nutrient stress conditions and shifts to competition under adverse conditions.In this study,pLERwheatdecreased due to increasing N application rates and pLERfababeandeclined due to increasing P application rates (Fig.6-A and E).These results indicated that the facilitation of N acquisition benefited the total intercropping yield and especially favored IW yield under N stress conditions.Moreover,N and P application rates decreased facilitation and total LER (Fig.6-C).Evidently,many factors,including sowing density,relative sowing time,and fertilizer application,affect companion crop interactions in intercropping and regulate its yield advantages (Liet al.2014;Yuet al.2016).Consequently,the regulated interactions between wheat and faba beans by optimized fertilizer application is a good strategy for optimizing intercropping economic and ecological functions.

4.3.N and P management in cereal and legume intercropping systems

N management in legume-based intercropping systems is important due to N-related nodulation and symbiotic N2fixation in legumes (Liet al.2009;Pelzeret al.2012).Clearly,balancing N input and output and the stimulated conversion of atmospheric N2into crop available N could collectively maximize the productivity and ecosystem service functions of legume-based intercropping systems (Huet al.2017).Previous studies have reported that legume-based intercropping reduced N fertilizer use,but not crop yield(Gaudinet al.2014;Layeket al.2014).Similarly,the optimal N application rate (75 kg ha-1) for forage yield was established in temperate regions for W//F,which was lower than other forage crops (Ghanbari-Bonjar and Lee 2002).In this study,W//F also exhibited the potential to save N fertilizer and enhance crop yield.Specifically,applying only half the N rate to IW did not result in yield losses compared to MW (Fig.2).It was estimated that wheat and faba beans could save 40-50% of N fertilizer and sustain similar wheat yields as mono-cropping based on the model analysis(Fig.8-A and B).Moreover,the optimal N application rate of IW was 150 kg N ha-1,which was lower than MW,and the optimal IW yield was 15% higher than MW (Fig.8-A and B),which was likely due to W//F improving the NUE values(Chenet al.2017;Xiaoet al.2018).Additionally,the current N application rates in wheat production were high and the optimal N application rate for MW was found to be 120-170 and 110-150 kg ha-1for IW.

Previously,cereal and legume intercropping were found to improve soil P availability and enhance P acquisition (Liet al.2007;Heet al.2013).Chenet al.(2017) argued that rational P management strategies contributed to enhanced total grain yields and P use efficiency in a relay strip intercropping system.When wheat was intercropped with faba beans,soil microbial functions and mycorrhizal soil infectivity were promoted (Wahbiet al.2016a,b),and both enhanced P uptake,N2fixation,and N transfer (Liet al.2016;Ingraffiaet al.2019).In this study,decreased P fertilizer(30-40%) use was maintained in IW,which was similar to MW yield (Fig.6-C and D).Hence,50-60 kg P2O5ha-1is recommended as the optimal dose for wheat production based on the results of both five-year field experiments(Fig.8-C and D).

Faba beans maintain relatively high biological N fixation levels,even in the presence of soil mineral N (Schwenkeet al.1998),which is more suitable as green manure in intercropping when N fertilizer application is required (Roseet al.2016).In this study,higher faba bean yields were obtained when the N rate was 45 kg ha-1as compared to other N rates.Thus,suitable faba bean N rates should be taken into account in intercropping N management practices;however,further investigations should be conducted in the future to verify the findings of this study.

5.Conclusion

W//F exhibited yield advantages (LER=1.12) due to stimulated IW yield,and intercropping insured against crop failure particularly under low N input.A negative relationship between wheat and faba bean was observed in intercropping.pLERwheatdecreased due toNapplication rates and pLERfababeandecreased due to P application rates,which occurred because the intercropping yield advantages decreased as the N and P application rates increased.Therefore,optimal N and P application rates could be a good strategy for maximizing intercropping yield and economic effects.Under the conditions of this study,IW yield was similar to that of MW even with decreased N input by 40-50% and P input by 30-40%.Application rates of 150 kg N ha-1and 60 kg P2O5ha-1for wheat are recommended for maximizing the productivity of W//F in southwestern China and similar growing regions.Thus,the nutrient management of faba beans should be considered in intercropping,and further investigations on the recommended application rates are needed for these practices.

Acknowledgements

This work was supported by the National Key R&D Program of China (2017YFD0200200 and 2017YFD0200207),the National Natural Science Foundation of China (31760611,32060718 and 31560581) and the Yunnan Agricultural Foundation Joint Project,China (2018FG001-071).We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

Declaration of competing interest

The authors declare that they have no conflict of interest.