The field experiments were performed in six winter wheat-summer maize rotations from 2010 to 2016 at Henan Agricultural University Experimental Station, located in the west of the North China Plain (113°38′39″ E, 34°47′51″ N). The soil is an alluvial soil (Gong et al., 2007), belonging to the Ochri-Aquic Cambosols (32.3% sand, 57.7% silt, 10.0% clay). In the cultivated layer (0.2 m), the pH was 7.82 with a field capacity of 23.9% volumetric soil water and a bulk density of 1.28 g m⁻³. The amounts of organic matter, total N, available phosphorus and available potassium were 17.47 g kg⁻¹, 0.84 g kg⁻¹, 18.83 mg kg⁻¹, and 252.56 mg kg⁻¹, respectively. The experimental site is characterized by a north temperate continental monsoon climate zone with an annual mean temperature of 14.3°C and annual sunshine duration of 2300 h. The average (1951–2006) annual rainfall is 617.1 mm, with the majority (60–75%) occurring in the summer months (June to September), and an average of 201.2 mm falling during the wheat growing season. Precipitation during the experimental seasons for wheat and maize is shown in Figure 1. Relatively dry weather prevailed during the wheat growing seasons of 2010–2011, 2012–2013, and 2015–2016; rainfall was 84.8, 183.9, and 194.3 mm, respectively, compared to the long-term average of 201.2 mm (1951–2006). However, the rainfall pattern showed monthly variability. During the 2012–2013 and 2015–2016 wheat growing seasons, there was only 13.1 and 7.9 mm of rainfall from jointing to flowering (March 10 to April 15), which was a major limiting factor to wheat production (Xue et al., 2006; Karam et al., 2009).

The experiment, a typical winter wheat-summer maize rotation, started with the winter wheat season of October 2010, and ended after the winter wheat harvest in June 2016, covering 11 successive seasonal crops in the same experimental plot. Each experimental plot was 7 m in length and 2.9 m in width. The plots were arranged in a split-plot design with four replicates. The main plots were assigned to three irrigation regimes: no irrigation water applied after sowing (W0), a single irrigation at jointing (W1), and two irrigations, one each at jointing and flowering (W2) during the wheat growing seasons. During the maize growing seasons, in all treatments irrigation was provided based on critical growth stages, such as the V12 stage, tasseling stage, and elongation stage during the 2012, 2013, and 2014 maize growing seasons, respectively (Table 1). Irrigation was applied by a movable sprinkler system, and 75 mm was applied each time as measured by a water meter. The treatment details are given in Table 1. Subplots were treated with five nitrogen rates: 0, 90, 180, 240, and 300 kg N ha⁻¹ during wheat growing season, and 0, 112.5, 225, 300, and 375 kg N ha⁻¹ during the maize growing season; these are referred to as N0, N1, N2, N3, and N4, respectively.

Wheat (cv. “Aikang 58” until 2012, then “Yumai 49–198”) was sown on 12 October each year at a sowing rate of 225 kg ha⁻¹ with 0.2 m row spacing. The wheat was harvested in late May-early June of the following year. Phosphorus was applied as triple superphosphate (P₂O₅; 46%) at a rate of 150 kg P ha⁻¹, and potash was applied as potassium chloride (K₂O; 46%) at a rate of 100 kg K ha⁻¹. Half the amount of nitrogen (urea; 46% N), and all phosphorus and phosphate was spread uniformly over the soil surface and then drilled into the upper 0.15 m at sowing. The other half of the nitrogen was applied at the jointing stage for each treatment. Maize (cv. “Zhengdan 958”) was planted by hand immediately after the wheat harvest in the same plots. P₂O₅ was applied at 90 kg ha⁻¹, K₂O at 120 kg ha⁻¹, and 30% of the nitrogen for each treatment was applied at the elongation stage of maize, with the remaining 70% applied at the V12 stage.

Additional protective measures were taken to ensure healthy growth such as the spray application of insecticides, fungicides, and herbicides. Accordingly, no significant incidence of pests, diseases, or weeds was observed in any treatment during any of the growing seasons.

Soil cores were collected every 0.2 m to a final depth of 1.6 m using a soil auger (0.05 m diameter) before sowing and after harvest in all experimental plots. Two cores were collected from each plot. Samples were then analyzed for soil NO₃-N content (SNC; mg kg⁻¹). Briefly, samples of fresh soil were extracted by shaking with a 2 M KCl solution (soil: solution ratio: 1:5) for 1 h. After filtering, the extract solution was analyzed using a continuous flow analyzer. Soil NO₃-N accumulation (SNA) was then calculated using the following formula:

Where, ρ_b (g cm⁻³) is the soil bulk density, and h (cm) is the soil thickness. The gravimetric water content (θ, g g⁻¹) of the samples was determined by drying them to a constant weight in an oven at 105°C (Bao, 2000). Soil water storage (SWS, mm) was calculated as:

In addition, soil water storage reduction amount (SWR) was calculated as the SWS at sowing minus the SWS at maturity for the different soil layers.

The evapotranspiration (ET) of the wheat was calculated by the following formula:

where ET (mm) is evapotranspiration, I (mm) is the amount of irrigation water that was measured using the water meters, P (mm) is the precipitation that was measured with the weather station at the site shown in Figure 1, ΔS (mm) is the SWS in the soil layer from 0 to 1.6 m at sowing minus the SWS at harvest, R (mm) is surface runoff (there was no surface runoff, so this variable was ignored). U is the upward capillary flow into the root zone (mm), and D is the downward drainage out from the root zone (mm). Because the groundwater table was ~10.2–11.2 m below the soil surface during the 6-year period, capillary flow is negligible (Liu and Wei, 1989). Drainage from the root zone was estimated using the following equation:

Where K (θ) is the unsaturated water conductivity when soil moisture is θ (cm day⁻¹), ΔH is the soil water potential between the top-soil and sub-soil (cm), ΔZ is the soil thickness (cm), and Δt is the period of time (in days). SWC was measured in the 1.6 m soil profile in this study, since more than 90% of the root system of wheat plants occupies only the first 1.0 m soil layer in the North China Plain, as was shown by several previous studies (Zhang et al., 2009; Wang et al., 2014; Xu et al., 2016). Therefore, drainage out from the root zone can be ignored in the North China Plain, including at our experimental site (Lv et al., 2011). Water use efficiency (WUE; kg ha⁻¹ mm⁻¹) in wheat is defined as grain yield divided by ET over the whole wheat growing season.

Roots were sampled at maturity by removing soil blocks during the 2013–2014 wheat growing season. Briefly, blocks were dug at 0.2 m intervals to a vertical depth of 1.6 m in six separate layers. Each sampling area was 0.4 m in length (perpendicular to the rows, providing access to plants in two rows) and 0.4 m in width (parallel to the rows). The roots with soil were then transferred to a 100–mesh nylon bag and submerged in water for 1 h. Any remaining soil was then washed from the roots using a low-pressure garden hose, and the clean roots were transferred to a sieve (0.25 mm² mesh) suspended in a trough partially filled with water. The cleaned roots were imaged using a scanner for the gray-scale scanning (Epson perfection V700 photo). The files were then analyzed using WinRHIZO 2008 in order to measure root length. Root length density (RLD; mm cm⁻³), the unit volume of soil root length, was determined using the following formula:

where L is the total root length (mm) and V is the volume of the soil sample (cm³).

Statistical analyses were performed using SPSS 17.0 software. Irrigation regime, nitrogen rates, and their interactions were taken as fixed factors, while the growing season was considered to be a random factor due to unpredictable weather conditions. Analysis of variance (ANOVA) was conducted for grain yield, SWS and SNA, ET and WUE, and for SWS, SNC, and RLD at each soil layer. The means of each treatment were compared with Duncan's multiple range test at the 0.05 probability level (at p ≤ 0.05). Pearson correlations were used to analyze the relationship between RLD and SWC (n = 72) and SNC (n = 36), and between yield and water-nitrogen use parameters (n = 144). Figures were prepared using Sigmaplot 12.3.

Wheat yields averaged from 2055 to 8690 kg ha⁻¹ in all treatments during the six growing seasons (Figure 2). Compared with the non-irrigated treatment (W0), irrigation at jointing (W1) significantly increased yield by 25.7, 3.6, 31.8, 13.6, 5.6, and 45.6% in the growing seasons from 2010–2011 to 2015–2016, respectively (Figure 2, Table 2). With increased irrigation (i.e., W2), wheat yield was significantly increased by 14.1, 45.3, and 29.3% compared to W1 in the 2010–2011, 2012–2013, and 2013–2016 growing seasons, respectively, in which the rainfall was lower or unevenly distributed during the wheat growing seasons (Figure 1). Correlation analysis (Table 5) revealed a strong positive correlation between rainfall and wheat yield (r = 0.601^**). However, no significant differences were found between W1 and W2 in the 2011–2012, 2013–2014, and 2014–2015 growing seasons (Figure 2, Table 2). This indicated that overfull irrigation did not result in higher yields in normal precipitation growing seasons. Wheat yields in irrigated treatments varied less over time compared to non-irrigated treatments, indicating that irrigation not only increased but also stabilized wheat yield. The CV for wheat grain yield in W0 was higher than in the irrigated treatments (W1 and W2), indicating that the precipitation and its distribution may be other important factors affecting wheat yield (Figure 1).

N application significantly improved wheat yield, but excessive N rates did not have a positive effect on yield (Figure 2). In the growing seasons from 2010–2011 to 2015–2016, the wheat yield (averaged across irrigation and N treatments) was increased by 3.6, 44.7, 78.5, 88.5, 122.6, and 76.8% in the N treatments compared to the N0 control treatment, respectively (Figure 2, Table 2), suggesting that the effect of N fertilizer on wheat yield became greater as the experimental seasons progressed. Averaged across the six growing seasons, the highest yield of 6711.3 kg ha⁻¹ was in the N3 treatment, but there was no significant difference between N2 and N3. Statistical analysis showed that the interaction between irrigation regimes and N rates on yield was significant (Table 2). The W2N3 treatment produced the maximum yield, and the W0N0 treatment gave the lowest yield (Figure 2).

Maximum SWS (averaged across irrigation and N treatments) before sowing of wheat (388.9 mm) was recorded in the 2011–2012 growing season, whereas the minimum SWS of wheat (274.9 mm) occurred in the 2012–2013 growing season (Table 3), which resulted in the highest and the lowest grain yields in the corresponding seasons (Table 2). No difference in SWS before sowing was observed among the three irrigation regimes in the 2013–2014 and 2014–2015 growing seasons, which resulted in reduced yield increases in the irrigated treatments (Table 2). SWS before sowing in N0 was higher than in the N treatments, and irrigation therefore did not increase yield under no-N conditions. N3 had the lowest SWS before sowing, while W2 gave the highest yield (Figure 2). This suggests that irrigation can increase yield under N application conditions, but that it could not balance the adverse effect of N stress on grain yield under conditions of N deficiency.

Plots of the soil water reduction amount (SWR) in the different soil layers are shown in Figure 3. Water was mainly consumed in the 0.4–1.2 m deep layers, with the highest SWR in the 0.6–1.0 m layer. The proportion of SWR in the 1.0–1.6 m layer was higher in the drier growing season 2012–2013 (32.0%) than in the normal precipitation growing season 2013–2014 (22.4%). Similarly, SWR in W0 was higher than in W1 and W2 at all soil layers (Figure 3). N rates had a significant effect on SWR, since N application increased SWR in all soil layers. The N3 treatment gave the highest SWR (Figure 3). Moderate water applied (W1) and appropriate N rates facilitated longer root growth, which resulted in more water being consumed in the deeper soil layers.

Irrigation caused a marked increase in evapotranspiration (ET). Compared to the un-irrigated treatments, ET in the W2 treatment increased by 45.2% during the drier growing seasons (averaged across 2012–2013 and 2015–2016), but only increased by 18.6% during growing seasons with normal precipitation (averaged across 2013–2014 and 2014–2015; Table 3). The W2 treatment showed a significant reduction in WUE in the 2013–2014 and 2014–2015 growing seasons, but gave the maximum WUE values during the 2012–2013 and 2015–2016 growing seasons (Table 4), which might be attributed to less rainfall and uneven distribution (Figure 1). N fertilizer caused an increase in ET that averaged 32.5, 38.2, and 34.5 mm for the N180, N240, and N300 treatments compared to N0, respectively (Table 3). Compared to the N0 treatment, N treatment significantly increased WUE by 76.4, 69.0, 96.1, and 70.9% (averaged across irrigation and N treatments) for the 2012–2013 to 2015–2016 growing seasons, respectively (Table 4). However, there was no further increase in WUE when the nitrogen application rate exceeded 180 kg N ha⁻¹.

W2 gave the lowest soil NO₃-N content (SNC), while W0 had the highest SNC in the top 0.4 m soil layer. However, from 0.8 m downwards, SNC in the irrigated treatments was greater than in the W0 control (Figure 4). This suggests that irrigation increased the levels of residual NO₃-N in the deep soil layers. Compared to N0, all N treatments significantly increased SNC in the 0–1.6 m soil layer at both flowering and maturity (Table 3, Figure 4). SNC in the N3 treatment was higher than in N2 in the 0.6–1.4 m soil layers at flowering, while no difference was found between N2 and N3 at maturity, indicating that N3 improved absorption of NO₃-N in the 0.6–1.4 m soil layers during the later stages of growth. This might be attributed to plants in the N3 treatment having the most extensive root systems in the deep soil layers. SNC in the 0.8–1.6 m soil layers in the N4 treatment was 37.7 and 86.9% higher than in N3 at flowering and maturity, respectively (Figure 2), implying that there was more residual NO₃-N in the deep soil layer in N4, which exceeded the ability of wheat to absorb it. Table 3 also shows that the soil nitrogen supply in N0 gradually declined as the experimental seasons progressed, which was in agreement with the slight decrease observed in grain yield in the N0 treatment (Figure 2).

Root length density (RLD) declined with increasing soil depth. Averaged across the irrigation and N treatments and both growing seasons, the highest RLD was 21.0 mm cm⁻³ in the 0–0.2 m soil layer, gradually decreasing to 10.1, 6.0, 4.9, 3.7, 3.0, 2.5, and 2.2 mm cm⁻³ at 0.2 m intervals through the soil column to a depth of 1.6 m. The W2 treatment had the highest RLD, while W0 plants had the smallest root system in the top 0.4 m soil layer (Figure 5). The RLD in the W0 and W1 treatments was greater than in W2 from 0.8 m downwards. The N application rates of 180 and 240 kg ha⁻¹ (N2 and N3) gave higher RLD values at depths from 0.4 to 1.6 m, while a higher RLD was only found in the 0–0.4 m soil layer in the N4 treatment. The relationships between RLD distribution and SWR and SNC for the different soil profiles were analyzed at 0.4-m intervals (Table 6). Positive relationships between RLD and SWR, and RLD and SNC were found in each soil layer for the non-irrigated and no-N treatments. SWC and SNC at soil depths of 0–0.4 and 0.8–1.6 m exhibited significant positive correlations with RLD for the W1 and N2 treatments, while a significant correlation was only found in the upper soil profile (0–0.4 m) for the W2 and N4 treatments. This suggests that elevated water and N use in deep soil layers under optimum irrigation and N regimes can be attributed to a deeper root system.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.