Suitability of the DNDC model to simulate yield production and nitrogen uptake for maize and soybean intercropping in the North China Plain

ZHANG Yi-tao , LlU Jian, WANG Hong-yuan LEl Qiu-liang LlU Hong-bin ZHAl Li-mei REN Tian-zhi, ZHANG Ji-zong

1 Key Laboratory of Non-point Source Pollution Control, Ministry of Agriculture, Beijing 100081, P.R.China

2 Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, P.R.China

3 Earth Systems Research Center, Institute for the Study of Earth, Oceans and Space, University of New Hampshire, New Hampshire 03824, USA

4 School of Environment and Sustainability, Global Institute for Water Security, University of Saskatchewan, SK S7N 0X4, Canada

5 Agro-Environmental Protection Institute, Ministry of Agriculture, Tianjin 300191, P.R.China

Abstract Intercropping is an important agronomic practice. However, assessment of intercropping systems using fleld experiments is often limited by time and cost. In this study, the suitability of using the DeNitriflcation DeComposition (DNDC) model to simulate intercropping of maize (Zea mays L.) and soybean (Glycine max L.) and its aftereffect on the succeeding wheat(Triticum aestivum L.) crop was tested in the North China Plain. First, the model was calibrated and corroborated to simulate crop yield and nitrogen (N) uptake based on a fleld experiment with a typical double cropping system. With a wheat crop in winter, the experiment included flve treatments in summer: maize monoculture, soybean monoculture, intercropping of maize and soybean with no N topdressing to maize (N0), intercropping of maize and soybean with 75 kg N ha–1 topdressing to maize (N75), and intercropping of maize and soybean with 180 kg N ha–1 topdressing to maize (N180). All treatments had 45 kg N ha–1 as basal fertilizer. After calibration and corroboration, DNDC was used to simulate long-term (1955 to 2012)treatment effects on yield. Results showed that DNDC could stringently capture the yield and N uptake of the intercropping system under all N management scenarios, though it tended to underestimate wheat yield and N uptake under N0 and N75.Long-term simulation results showed that N75 led to the highest maize and soybean yields per unit planting area among all treatments, increasing maize yield by 59% and soybean yield by 24%, resulting in a land utilization rate 42% higher than monoculture. The results suggest a high potential to promote soybean production by intercropping soybean with maize in the North China Plain, which will help to meet the large national demand for soybean.

Keywords: maize intercropping with soybean, DNDC, topdressing N, yield, N uptake

1. lntroduction

The North China Plain is very important to China’s food security, producing nearly 20% of the national grain crops(NBSC 2016). Dominated by rotation of maize (Zea maysL.)-wheat (Triticum aestivumL.) crops, crop production in the North China Plain is greatly intensifled with large plant densities and high inputs of irrigation water, chemical fertilizers, and pesticides (Fanget al.2006, Chenet al.2014; Juet al.2016). However, excessive applications of chemicals have caused severe environmental problems (Zhanget al.2004) including nitrate pollution of groundwater (Juet al.2006), greenhouse gas emissions(Zhanget al.2012), and soil acidiflcation (Blumenberget al.2013). At present, nitrogen (N) loss is one of the main concerns in this region. Large amounts of nitrate leaching are observed during maize growing seasons, as a result of heavy summer rains combined with excessive N application, a common phenomenon in the North China Plain (Juet al.2009). Therefore, best management practices are necessary to ensure both agronomic productivity and environmental quality. Potential best management practices lie in sustainable cropping systems that can efflciently utilize solar radiation and land resources with minimal anthropogenic inputs (Zhanget al.2015a).For example, an intercropping system is demonstrated to have advantages of yield increase and improved light and heat utilization over crop monocultures (Zhang and Li 2003; Liuet al.2017).

An intercropping system usually contains two or more crops grown simultaneously for a certain period of time(Zhanget al.2007a), so at least two crops can be harvested in one growing season while maintaining the yield of the main crop. Often, intercropping leads to high productivity, effective control of pests and diseases, efflcient resource utilization,good ecological services, and better economic beneflts(Thierfelderet al.2012; Xiaet al.2013; Midegaet al.2014;Wu and Wu 2014). As involvement of more crops can result in higher labor costs, in practice only two crops are used in most intercropping systems (Cavigliaet al.2011). Moreover,challenges remain for sowing and harvesting crops as well as weed control in intercropping systems (Feikeet al.2010).Therefore, an optimal intercropping pattern, which is suitable for mechanical operation, is required to compensate for the complexity of fleld management and labor costs. Among different intercropping patterns, strip intercropping, i.e.,one crop strip intercropped with another crop strip, is most convenient for mechanical operations (Lesoing and Francis 1999). Cereal intercropping with legumes is one of the most popular options. In a previous study in the North China Plain,Zhanget al.(2015a) successfully intercropped soybean with maize in a strip intercropping system and determined that the optimum ratio of rows of maize and soybean was 4:6,because it allowed for machine operations.

Intercropping soybean with cereal crops has very important agronomic and environmental implications. In China, the yield of soybean is commonly low but the price is very high, and the imported, genetically modifled soybean is not well accepted by consumers (Zheng and Wang 2013;Wang and Zhu 2016). The main region growing soybean is located in Northeast China, however, the growing area has decreased due to low yield during past years (Iizumi and Ramankutty 2016). To promote soybean production,China’s government aims to increase the soybean growing area to 9.3 million hectares by 2020, an increase of 2.7 million hectares from 2015 according to the Guidelines for Promoting the Development of Soybean Production issued by the Ministry of Agriculture of China. Intercropping soybean during the maize growing season is one of the most promising practices to improve soybean planting area without reducing maize yield. Moreover, maize intercropping with soybean could reduce N use in per unit area compared to maize monoculture, because soybean can flx N in the atmosphere and thus requires little additional fertilizer N inputs, if any (Moyer-Henryet al.2006; Yanget al.2015;Zhanget al.2015a). Nitrogen flxed by legumes could be transferred to maize, which could further increase maize yield and reduce N application to maize strips (Fanet al.2006). Higher yield can result in more N uptake from soil.Additional N uptake advantages in intercropping systems can be derived from enhanced light use efflciency aboveground and enhanced nutrient (e.g., N) use efflciency belowground (Lvet al.2014). Moreover, inclusion of legumes in intercropping systems was demonstrated to have positive aftereffect, i.e., beneflting yield production of subsequent crops (Olasantan 1998; Bergkvistet al.2011; Zhanget al.2015a).

Use of crop models is an important approach to analyzing yield potentials, yield gaps, or N utilization.One of the greatest strengths of model simulations is that verifled models can accurately predict yield variabilities and examine long-term effects of different planting patterns using available weather data (Chenet al.2015; Zhanget al.2017). Indeed, some models can be used to simulate intercropping systems. These models often include competition for light, water, and N, such as ALMANAC(Agricultural Land Management Alternatives with Numerical Assessment Criteria) for weed relay intercropping with wheat (Debaekeet al.1997), STICS (Simulateur mulTIdisciplinaire pour les Cultures Standard) for pea–barley intercrops (Corre-Hellouet al.2009), RUE (radiation use efflciency) for wheat-maize relay strip intercropping(Gouet al.2017), and FASSET (Farm Assessment Tool) for pea intercropped with spring barley (Berntsenet al.2004).However, the existing models are generally established for full mixtures of intercropping crops and are less suitable for strip intercropping (Gouet al.2017). In contrast, the most well-known crop models used for research such as DNDC(DeNitriflcation-DeComposition) and DSSAT (Decision Support System for Agro-technology Transfer) mainly focus on crop monocultures (Songet al.2009; Minet al.2011;Zhanget al.2015b). Whether such common models could be used to simulate intercropping systems has not been reported. Because these models require relatively fewer parameters, they can be potentially widely used if proven to be suitable for simulating intercropping systems.

In this study, we focused on the DNDC model, because it is a biogeochemical model that is process-based and it has been verifled by fleld data worldwide (Tonittoet al.2007;Denget al.2011). The model combines biogeochemical processes with hydrological dynamics, and it can be used to simulate physiological processes such as N uptake, N stress, and water stress during growth of various plants(Zhanget al.2002). A large number of studies have used DNDC to identify the best management practices to achieve yield or environmental goals (Gopalakrishnanet al.2012;Werneret al.2012). However, the model is usually used to simulate carbon and N cycles of monoculture systems(Liet al.2014; Zhanget al.2015b), and it has not been tested for intercropping systems. Based on its applicability to monoculture simulations, we simulated two crops planted simultaneously to form an intercropping system,and we assumed that different intercropping patterns were reflectedviathe maximum biomass of two crops. Results of competition for resources aboveground (e.g., light) and belowground (e.g., water and nutrients) could be shown by crop yield. In this study, we calibrated and corroborated the DNDC model based on fleld experiments that examined crop yield and N uptake as affected by different N application rates for maize and soybean intercropping as well as maize monoculture and soybean monoculture and used the model to simulate the long-term effects of intercropping and N management. The objectives were to: (1) evaluate the applicability of the DNDC model to simulate yield and N uptake for intercropping systems under different N application rates, and (2) identify advantages of intercropping compared with monoculture based on longterm DNDC simulations.

2. Materials and methods

2.1. Study area

2.2. Field experiments

Experimental designThis study was based on the traditional rotation of summer maize and winter wheat(maize-wheat), for which soybean was introduced to be intercropped with maize. The experiment was conducted with a randomized complete block design from June 24th,2011 to June 17th, 2012. It included flve treatments with different crop and N management practices in the summer growing season but identical management for winter wheat.The different summer treatments were: monoculture of maize, monoculture of soybean, and intercropping of maize and soybean with three different N treatments, where maize always received 45 kg N ha–1basal urea fertilizer(46% N) but with 0 N ha–1(N0), 75 N ha–1(N75), or 180 kg N ha–1(N180) topdressing urea fertilizer at the jointing stage (i.e., 40 days after sowing). Monoculture of maize,monoculture of soybean, and winter wheat were supplied with 225, 45 and 225 kg N ha–1as urea, respectively. For soybean, all N fertilizers were applied basally, while for monoculture of maize and winter wheat, half of the N was incorporated into the top 20 cm soil as a base fertilizer at sowing and the remaining half was topdressed 40 days after planting. All crops were supplied with 33 kg P ha–1in calcium superphosphate (12% P) and 62 kg K ha–1as potassium sulfate (52% K) as basal fertilizers. Topdressing was implemented by broadcasting urea on the soil surface between maize rows. Each treatment was replicated three times in plots that have an area of 70 m2for monoculture of maize or soybean and 110 m2for intercropping of maize and soybean (Table 1).

The detailed planting pattern of the experimental cropping systems is shown in Table 1. For intercropping of maize and soybean, four rows of maize were intercropped with six rows of soybean as recommended by Zhanget al.(2015a).Both maize (variety Zhengdan 958, Henan Academy of Agricultural Sciences) and soybean (variety Zhonghuang 30,Chinese Academy of Agricultural Sciences) were sown on June 24th, 2011 and harvested on October 6th, 2011. After harvest of summer crops, a subsequent winter wheat (variety Tangmai 6, Tangshan Academy of Agricultural Sciences)crop was sown to all plots (at a 15-cm inter-row distance)on October 7th, 2011 and harvested on June 17th, 2012.Throughout the yearly crop rotation, six flooding irrigations were applied: 40, 60, 60, 70, 70, and 70 mm of water at maize growth stages of sowing, and wheat growth stages of sowing, overwintering, erecting, booting and fllling,respectively.

Sampling and analysis of soil and plantsBefore plants were sown on June 24th, 2011, soils were sampled randomly across the experimental fleld using a 5-cm diameter soil auger to determine basic physical and chemical properties. Soil moisture was measured with the oven-drying method: a total of 20 g fresh soil was dried at 105°C for 12 h to a constant weight. Soil organic matter and total N concentrations were determined by the potassium dichromate method and the automatic Kjeldahl method,respectively, after wet digestion (Perezet al.2001).

At harvest, samples of maize, soybean, and wheat were collected from each plot to determine yield and N uptake.Ten plants of maize or soybean were sampled from every row in summer, and wheat plant samples within an area of 2 m2were taken separately for the previously established maize or soybean strips. Stalks and grains were harvested separately. Plant samples were oven-dried flrst at 105°C for 30 min and then at 85°C to a constant weight. After grinding, the samples were wet-digested with H2SO4and H2O2to determine total N content by the automatic Kjeldahl method (Lithourgidiset al.2011).

2.3. Model simulations with DNDC

Model descriptionOriginally, the DNDC model was developed to simulate nitrous oxide emissions from agricultural lands in the United States (Liet al.1992a, b).During the past 20+ years, the model was expanded to assess C and N turnover in agro-ecosystems, involving the processes of methane emissions, ammonia volatilization,changes in soil organic carbon and soil climate, crop production, and nitrate leaching, etc (Liet al.1997, 2005,2006; Li 2000; Denget al.2011; Zhanget al.2015b). Using the inputs of ecological drivers such as meteorological data,soil properties, and crop and nutrient management practices,DNDC simulates soil environments such as temperature,moisture content, oxidation-reduction potential, pH, and substrate concentration gradients. Furthermore, DNDC simulates crop growth and turnover of nutrients including microbial related processes of nitriflcation, de-nitriflcation and fermentation.

In the DNDC model, there is a component to simulate crop growth, for which photosynthesis, respiration, C allocation, water and N uptake by crops are calculated on a daily basis during simulation. Both water and N uptake rely on several factors such as soil N distribution, soil moisture content, and root length, etc. Water utilization depends on potential transpiration linked with leaf area index and climatic conditions. Water stress is simulated when potential transpiration is relatively higher than normal or actual water supply. Nitrogen need by crop could be calculated according to the optimal crop growth and plant C/N ratio on a daily basis, while plant growth would be inhibited by N stress when plant N uptake is below a critical value.

Model setup, calibration and corroborationIn this study,we used the newest model version (DNDC95) downloaded from the internet (UNH 2013). The DNDC model was set up with relevant inputs including daily meteorological data(the maximum and minimum temperature, precipitation,wind speed and humidity), atmospheric N deposition and NH3concentration, soil information (texture, pH, bulk density and SOC content), and fleld management practices of crop, tillage, fertilization and irrigation. During the simulation process, the 50-cm soil proflle was characterized as most of the crop roots were concentrated from 0 to 50 cm. In order to demonstrate the applicability of DNDC to simulating intercropping, the model was flrst calibrated with the N180 treatment of intercropping of maize and soybean then rotation with wheat. Simulated crop yield and N uptake were compared with the measured values. Comparisons of default and adjusted values (based on measurements or previous studies (Li 2007)) for soil and crop parameters are shown in Tables 2 and 3, respectively. After calibration, the DNDC simulated crop yield and N uptake were corroborated with the observed values in the treatments of N0 and N75 in the cropping system of maize intercropped with soybean then rotation with wheat. Finally, the verifled model was used to simulate long-term yield production in different cropping systems with climatic data from 1955 to 2012. Taking the flrst flve years from 1955 to 1959 as the spin-up time, the objective of these simulations was to examine if summer intercropping had a yield advantage compared to traditional maize-wheat rotation over the long-term.

It had taken 40 years, but the good deed had been repaid. Nana was right. We reap exactly what we sow. “Every good deed you do will someday come back to you.”

Table 1 Detailed planting pattern of the experimental cropping systems

In the simulation process, parameters of max biomass production of intercropped maize and soybean were set up individually. The DNDC model could not separate N application to the intercropped crops, so maize and soybean were simulated with the same fertilization regime during their simultaneous growth period. This was different from the conventional farming practice that intercropped maize received N topdressing and intercropped soybean was only supplied with basal N. However, as soybean can flx atmospheric N and N is not a limiting factor, the N applied to soybean in the model would not affect its production and N uptake.

2.4. Data analysis

Land equivalent ratio (LER), which is usually considered as an indicator of intercropping beneflt (Tariah and Wahua 1985), was calculated according to:

Where,Yim(kg ha–1) andYis(kg ha–1) are respective yields of intercropped maize and soybean per ha intercropping area, andYsm(kg ha–1) andYss(kg ha–1) are yields of maize and soybean in monoculture. An intercropping system presents yield advantage over crop monoculture if LER is greater than 1.00.

In the fleld experiment, N uptake by a crop (Nup, kg ha–1) was calculated according to:

Where,Mis the crop dry matter (kg ha–1) at harvest andNconis the N concentration in the plant (%). Nitrogen uptake by grain and stalk was calculated individually and summed up to obtain total N uptake.

Table 2 Parameter adjustments for soil

?

3. Results

3.1. Model calibration of maize intercropping with soybean and its aftereffect on winter wheat

The DNDC simulated crop yields were strongly correlated with the observed yields of the intercropped summer maize and soybean and the succeeding winter wheat under the treatment of N180, in which maize received 45 kg ha–1basal N before sowing and was topdressed with 180 kg N ha–1at the jointing stage (Fig. 1-E). Speciflcally, the observed yields of the intercropped maize and soybean were (7 605±488) and(1 915±40) kg ha–1, respectively, and the yield of winter wheat was (7 408±330) kg ha–1. The simulated yields of these crops (7 345 kg ha–1for intercropped maize, 1 915 kg ha–1for intercropped soybean and 7 843 kg ha–1for wheat)were stringent (“very good”) for capturing the magnitudes of fleld observations.

Similar to yields, DNDC also performed very well in capturing the magnitudes of plant N uptake (Fig. 1-F).Simulated N uptake of intercropped maize and soybean were 122 and 130 kg ha–1, respectively, in comparison with the measured N uptake of (129±9) kg ha–1for intercropped maize and (127±3) kg ha–1for intercropped soybean.Simulated and measured wheat N uptake were 203 kg ha–1and (217±16) kg ha–1, respectively.

Fig. 1 Comparisons of measured and simulated yields and N uptake of maize, soybean and wheat under different treatments in the cropping system of maize intercropped with soybean then rotation with wheat. N0, 45 kg ha–1 basal N without topdressing;N75, 45 kg ha–1 basal N and 75 kg ha–1 topdressing N to maize; N180, 45 kg ha–1 basal N and 180 kg ha–1 topdressing N to maize.

3.2. Model corroboration of maize intercropping with soybean and its aftereffect on winter wheat

Comparisons between simulated and observed results demonstrated that the DNDC model stringently (“very well”) captured the magnitudes and patterns of yield and N uptake of intercropped maize and soybean for both treatments of N0 and N75 (Fig. 1-A–D). However, despite the accurate simulation of wheat yield by the model for N75, it underestimated the wheat yield of N0 and N uptake of both N0 and N75. Along with the comparisons for simulating N180, the results indicated that the model could predict yield production of all N management scenarios in the experimental of maize intercropped with soybean then rotation with wheat. The model could also accurately simulate N uptake of the intercropped crops but tended to underestimate N uptake by the succeeding crop, especially when a relatively low rate or no N was applied to the previous maize crop.

3.3. Simulation of monoculture of summer maize or soybean and its aftereffect on winter wheat

Notably, some of the parameter values obtained from simulating intercropping had to be recalibrated to simulate monoculture of summer maize or soybean. The maximum grain production of maize was adjusted to 4 500 for maize monoculture and even more parameters were adjusted for soybean monoculture to achieve yield and N uptake that were comparable to fleld observations. Speciflcally, the adjusted parameters for soybean monoculture included maximum biomass production, biomass fraction and biomass C/N ratio of grain, leaf, and stem, respectively(Table 4).

After calibration, crop yields in both maize-wheat and soybean-wheat rotations were accurately simulated by the model (Fig. 2). For maize-wheat rotation, the observed yields were (9 630±115) and (7 398±743) kg ha–1,respectively, compared to the simulated yields of 9 728 and 7 428 kg ha–1. For soybean-wheat rotation, the yields were (3 775±100) and (7 188±650) kg ha–1compared to the simulated yields of 3 903 and 8 058 kg ha–1. Similarly, N uptake of all crops except wheat in maize-wheat rotation was all accurately simulated (Fig. 2). Speciflcally, maize-wheat rotation took up (160±3) kg N ha–1in maize and (229±11)kg N ha–1in wheat in the fleld, while uptake of N by wheat was underestimated by 37 kg ha–1in the model simulation.The observed N uptake of soybean-wheat was (237±2)and (210±12) kg ha–1compared to 237 and 205 kg ha–1simulated by the model.

3.4. Long-term impacts of intercropping on yield and advantages of intercropping

For the simulations from 1960 to 2012, compared with monoculture of maize or soybean, intercropping patterns increased yields of both maize and soybean per unit area.Over the three N management scenarios, intercropping increased maize yield by 48–59% per hectare of maize growing area and increased soybean yield by 19–24%per hectare of soybean growing area. While the yield of intercropped soybean varied very slightly among the three N topdressing treatments ((2 630±625) kg ha–1yr–1in N0,(2 607±650) kg ha–1yr–1in N75, and (2 490±779) kg ha–1yr–1in N180), the yield of intercropped maize was signiflcantly affected by different N topdressing rates (Fig. 3). Given no N topdressing (N0), maize yield was as low as (3 268±179) kg ha–1, which was smaller than half of that with N topdressing of 75 kg ha–1((7 386±812) kg ha–1) and N topdressing of 180 kg ha–1((6 888±1 120) kg ha–1). Notably, N75 had even higher and more stable maize yield than N180. Therefore,a rate of 75 kg N ha–1could be considered as the optimal N topdressing rate for maize in the maize and soybean intercropping system.

With regard to the aftereffect, yield of winter wheat was affected by the previous crops and N management scenarios(Fig. 3). The highest wheat yield of (8 630±542) kg ha–1was achieved in the treatment with soybean monoculture compared to 7 373–8 418 kg ha–1in other treatments. The second highest wheat yield was found in the N180 treatment.The N0 and N180 treatments resulted in similar wheat yields,which were lower than that in maize monoculture.

Driven by precipitation, crop yields changed dramatically during the 53 years of simulation. For example, extremely low yields of maize and soybean were simulated in 1965, 1968, 1975, 1984, 1997 and 2000, where annual precipitations were only 307, 393, 207, 290, 301, and 242 mm, respectively. The results indicated that suitable water management needs to be adopted in addition to crop and N management to achieve high crop yields.

Table 4 Parameter adjustments for management of crop monoculture1)

Fig. 2 Comparisons of measured and simulated crop yields and N uptake in the cropping systems of maize-wheat (A and B) and soybean-wheat (C and D).

Evaluation on the land utilization rate (LER) showed that when appropriate N management was provided,intercropping of maize and soybean had a yield advantage over maize or soybean monoculture. The LER values of N75 (average 1.42 with range of 1.04–1.61) and N180(average 1.32 with range of 0.96–1.46) were both greater than 1, and the land utilization rates in these systems were 32–42% higher than the rates in the two monocultures.In contrast, the LER value of N0 (average 0.99 with range of 0.88–1.27) was lower than 1, indicating a lack of intercropping advantage.

4. Discussion

Generally, well managed intercropping could produce more yield than crop monoculture, which is an important measure to enhance food security. Historically, intercropping was widely used by farmers in China (Liet al.2001; Zhanget al.2007b), but its usage has declined in recent years with the increasing labor price and demand of suitable machineries (Feikeet al.2012). Optimally, intercropping should be designed to accommodate existing machineries so that farmers are willing to adopt this practice. In a previous fleld study, Zhanget al.(2015a) conflrmed that strip intercropping, which could both reduce labor inputs and be operated by plant seeders and harvesters, had a high potential for use in the North China Plain. However,uncertainties remain with regard to N management and weather effects on the efflciency of strip intercropping. While extensive evaluations are almost impossible to accomplish using fleld experiments due to limitations of time and costs,model applications can provide more efflcient evaluations.

Fig. 3 Comparisons of long-term crop yields in different crop and N treatments simulated with the DeNitriflcation-DeComposition(DNDC) model. N0, 45 kg ha–1 basal N without topdressing; N75, 45 kg ha–1 basal N and 75 kg ha–1 topdressing N to maize; N180,45 kg ha–1 basal N and 180 kg ha–1 topdressing N to maize.

This study conflrmed that the DNDC model could accurately simulate yield and N uptake of intercropped maize and soybean under different N management strategies.In the past, DNDC was mainly used to simulate C and N biogeochemistry in agro-ecosystems as well as yield production of crop monocultures (Zhanget al.2017). The conflrmation of DNDC’s ability to simulate intercropping systems in this study provides important insights to expand the application of this model. Use of DNDC could also potentially compensate for the weakness of many current models that are used to assess yield potentials or yield gaps in crop monocultures only (Lianget al.2011;van Ittersumet al.2013; Zhanget al.2015b). Notably,however, different values were needed for some crop parameters to best simulate intercropping and maize or soybean monoculture. This is most likely because crop growth patterns in intercropping are different from that in a monoculture due to interactions between the crops (Zhang and Li 2003; Zuo and Zhang 2008; Betencourtet al.2012).The model also accurately predicted yield and N uptake of wheat, which was grown following the intercropping, in the high N topdressing treatment (N180). However, the model tended to underestimate yield and N uptake of wheat when a low rate of N (N75) or no N (N0) were topdressed to the previous maize crop. This may be due to an underestimation of the aftereffect of intercropping on wheat growth by the model or an underestimation of atmospheric N deposition during the wheat growing season. In the future, the DNDC model should be further developed in the aspect of modeling the aftereffect of intercropping.

Despite fluctuation of yields with years due to weather conditions, long-term simulations showed constant yield advantage of intercropping over monoculture. The differences in crop yields between intercropping and monoculture may be partly due to different plant densities in the two systems. Notably, however, the plant densities implemented in this study were consistent with farmers’conventional practices. More importantly, the yield advantage of intercropping results from a balance between interspeciflc facilitation and competition (Zhang and Li 2003;Zuo and Zhang 2008; Betencourtet al.2012). Facilitative interactions, such as efflcient utilization of photosynthetically active radiation or higher radiation use efflciency (Liuet al.2017), can improve crop growth and nutrient utilization, while two crops completing for water and nutrients in intercropping systems could drive atmospheric N2flxation in soybean byRhizobium(Corre-Hellouet al.2006). The simulated results revealed that the higher rate of N topdressing (N180) did not achieve higher maize yield compared to that of the low N rate (N75). This may be an indication that there were fewer facilitative interactions in N180 due to sufflcient supply of N to maize. Obviously, however, N topdressing to maize is needed to produce high grain yield in intercropping because no N topdressing led to low yields of the intercropped maize and the succeeding wheat crop.

In the long-term simulation, intercropped maize and soybean with topdressing N of 75 kg ha–1to the system increased crop yield by 59 and 24% compared with monoculture of maize and soybean, respectively. The LER value of N75 was 1.42, greater than 1, which indicated that the land utilization rate was 42% higher than the rates of two monocultures. Moreover, the results suggest that yield of soybean could be increased through intercropping with maize in the North China Plain, which will help to meet the high consumer demand for soybean in China and alleviate reliance on importing soybean from other countries. It should be noted that crop yields in the low precipitation years are very likely to be underestimated in long-term simulations.In practice, farmers would adopt some additional measures such as irrigation or soil mulching to combat drought conditions. However, the simulation results seem to be valid because the same patterns were simulated under all weather conditions.

5. Conclusion

This study demonstrated that the DNDC model could be used to simulate yield production and N uptake in intercropping systems. The model stringently captured the yield and N uptake of intercropping maize and soybean under different N management scenarios in the summer.The model tended to underestimate yield and N uptake of the wheat crop grown after intercropping when intercropped maize received no or a low rate of topdressing N. This indicated a need to improve the ability of the DNDC model to simulate the aftereffect of intercropping. Long-term model simulations suggested that intercropping had greater agronomic efflciency than crop monoculture. Topdressing of 75 kg N ha–1to the intercropped maize produced higher crop yield than topdressing of 180 kg ha–1or no topdressing,providing application of the same rate of N (i.e., 45 kg N ha–1) as a basal fertilizer. Over the long term, intercropping maize and soybean could increase their yields by up to 59 and 24%, respectively, in per hectare growing area,compared to monoculture. Intercropping with N75 resulted in a LER value of 1.42. Furthermore, the results suggest a large potential to increase China’s soybean production by intercropping soybean with maize in the North China Plain.

Acknowledgements

This research was supported by the National Natural Science Foundation of China (31701995 and 31572208),the National Key Research & Development Program of China (2016YFD0800101), the Newton Fund of UK-China(BB/N013484/1). This paper was also supported by China Scholarship Council (2015-7169).