Effects of Different Fertilization Regimes on Crop Yield and Soil Water Use Efficiency of Millet and Soybean

Soil water and nutrients are major factors limiting crop productivity. In the present study, soil water use efficiency (WUE) and crop yield of millet and soybean were investigated under nine fertilization regimes (no nitrogen (N) and no phosphorus (P) (CK), 120 kg ha−1 N and no P (N1P0), 240 kg ha−1 N and no P (N2P0), 45 kg ha−1 P and no N (N0P1), 90 kg ha−1 P and no N (N0P2), 120 kg ha−1 N and 45 kg ha−1 P (N1P1), 240 kg ha−1 N and 45 kg ha−1 P (N2P1), 120 kg ha−1 N and 90 kg ha−1 P (N1P2), 240 kg ha−1 N and 90 kg ha−1 P (N2P2)) in the Loess Plateau, China. We conducted fertilization experiments in two cultivation seasons and collected soil nutrient, water use, and crop yield data. Combined N and P fertilization resulted in the greatest increase in crop yield and WUE, followed by the single P fertilizer application, and single N fertilizer application. The control treatment, which consisted of neither P nor N fertilizer application, had the least effect on crop yield. The combined N and P fertilization increased soil organic matter (SOM) and soil total N, while soil water consumption increased in all treatments. SOM and total N content increased significantly when compared to the control conditions, by 27.1–81.3%, and 301.3–669.2%, respectively, only under combined N and P application. The combined N and P application promoted the formation of a favorable soil aggregate structure and improved soil microbial activity, which accelerated fertilizer use, and enhanced the capacity of soil to maintain fertilizer supply. Crop yield increased significantly in all treatments when compared to the control conditions, with soybean and millet yields increasing by 82.5–560.1% and 55–490.8%, respectively. The combined application of N and P fertilizers increased soil water consumption, improved soil WUE, and satisfied crop growth and development requirements. In addition, soil WUE was significantly positively correlated with crop yield. Our results provide a scientific basis for rational crop fertilization in semi-arid areas on the Loess Plateau.


Introduction
Crop growth and development activities not only require appropriate light, moisture, air, and temperature conditions but also adequate amounts of diverse soil nutrients [1,2]. Considering limited soil nutrient availability, the application of fertilizer to the soil is necessary [3], which would also address current food security challenges by enhancing soil fertility and, in turn, crop yield. The increasing rates of fertilizer application in agricultural production over recent years, with generally low utilization rates, have led to the considerable waste of fertilizer resources, in addition to severe environmental pollution [4,5]. Controlled-release fertilization is an efficient and pollution-free method of fertilization, which addresses the low utilization rate challenge associated with ordinary quick-acting fertilizers, in addition to environmental pollution, and is a potentially effective strategy to ensure the sustainability of land use and efficient crop production [1,3].
Inadequate soil water and soil nutrient levels are the major factors limiting agricultural production in the Loess Plateau [6][7][8]. Agricultural production in such environments should aim at achieving efficient water use [9]. Due to the lack of irrigation water resources in the Loess Plateau, rain-fed agriculture accounts for a large proportion of the agricultural production in the region [5,10]. The considerable fertilizer inputs in the region have promoted crop yield greatly, in addition to the associated economic benefits, and the improvement of soil water use efficiency (WUE) of the cropping systems. Consequently, soil fertilization has become one of the key strategies of achieving higher crop yields in the region [11,12]. Rational fertilization practices that take into account soil water availability, improve soil WUE, and improve crop yield have become key strategies for comprehensive agricultural development on the Loess Plateau [13,14].
Strategies for improving crop yield and soil WUE through rational fertilization under poor water availability conditions have been extensively studied. Numerous studies have demonstrated that appropriate water and fertilization conditions facilitate nutrient absorption by crops which, in turn, facilitate coordinated growth in crops [15,16]. Particularly, nitrogen (N) and phosphorus (P) fertilization could decrease transpiration losses significantly and enhance WUE in crop production on the Loess Plateau [7].
Soil WUE reflects the relationship between crop production and water consumption. It is a comprehensive index for evaluating crop growth suitability, and improving WUE could minimize the consumption of limited water resources, especially in semi-arid areas where precipitation is scarce, with important ecological and economic benefits. In arid areas, the consumption of less water resources could promote sustainable water resource exploitation in ecosystems [3]. Drought stress in the Loess Plateau greatly influences fertilization efficiency in the region [17,18]. Under limited soil water conditions, water availability and N and P fertilization can be exploited collectively to enhance yield and improve crop quality [19]. Fertilization can promote crop growth and development, particularly by promoting the development of shallow roots, and enhancing soil nutrient absorption and utilization efficiency [20][21][22]. Due to crop cultivation and soil erosion, the soil organic matter (SOM) concentrations in dryland farmlands are generally low, and the ability to supply and maintain fertilizer is usually poor. Consequently, the efficient use of water and nutrients is critical for sustainable crop production on the Loess Plateau. However, only a few studies have investigated the effects of fertilization on soil WUE and yield on the Loess Plateau. In the present study, we established split-plot experiments in Ansai County, Shaanxi Province, China. The study region is in a semi-arid/semi-humid climate zone, and major soil and water conservation projects have been implemented in the area [7,8]. The aim of the present study was to investigate the effects of different fertilization regimes on crop yield and WUE on the Loess Plateau, and to determine fertilization strategies that could facilitate sustainable and efficient utilization of water and fertilizer resources in the arid areas of the Loess Plateau. We hypothesized that (1) fertilization significantly increases water use efficiency and crop yield, and (2) the combined application of nitrogen and phosphorus fertilizer has a greater positive effect on soil water use efficiency and crop yield.

Basic Geographic Information
The field experiments were conducted in 2017 and 2018 in the middle of the Loess Plateau (36 • 51 N, 109 • 18 E), at 1068 m above sea level, in an area with a mean annual temperature of 8.8 • C and a mean annual precipitation of 500 mm. The study site falls under the loess hilly and gully area. The soil type is Regosol and the soil nutrient and water loss levels in the area are high. Millet was planted in 2017 and soybeans were grown in 2018. In 2017, the total rainfall in the growth period was 443.1 mm, accounting for 79.46% of the total annual rainfall. In 2018, the total rainfall in the growth period was 503.1 mm, accounting for 93.8% of the total annual rainfall ( Figure 1). Sustainability 2020, 12, x FOR PEER REVIEW 3 of 14 mm, accounting for 79.46% of the total annual rainfall. In 2018, the total rainfall in the growth period was 503.1 mm, accounting for 93.8% of the total annual rainfall ( Figure 1). Basic soil physical and chemical properties in the study site were measured at a depth of 0-40 cm before fertilization (Table 1). Soil bulk density was measured using a soil bulk sampler and soil pH was measured using a pH meter. Notes: SOM = soil organic matter; TN = soil total nitrogen; TP = soil total phosphorus; all in (g·kg −1 ).

Experimental Design and Fertilizer Management
Local heat and moisture conditions dictate that crops are only grown once a year. Millet was planted in 2017 and soybeans were planted in 2018. The millet cultivar was Hongyang 7. The soybean variety planted was Zhonghuang 35. The order of nutrient absorption in millet was N > K > P. There were differences in nutrient concentrations among the millet crop organs, with P and K concentrations being the highest in the fruit, and the N concentrations being the highest in the root system. From the flowering period, the demand for N fertilizer began to increase rapidly to support grain growth. Nitrogen fertilizer could be applied at this period for optimal yield benefits. For every 100 kg of millet produced, 2-5.23 kg N, 0.2-0.75 kg P (P2O5), 1.0-1.98 kg K (K2O) are required. The order of nutrient absorption in soybean was N > K > P. Soybean stakes are more fattening crops, because the formation of nutrients such as protein and fat in soybeans requires high nutrient amounts, especially N, P, and K. For every 100 kg of soybean produced, 5.3-7.2 kg N, 1-1.8 kg P (P2O5), and 1.3-4.0 kg K (K2O) are required.
The experimental site is a rectangular block with an area of 567 m 2 , with nine fertilizer treatments, each with three replicates, in 3 × 7 m plots ( Table 2 and Figures 2 and 3). The N fertilizer was urea, while the P fertilizer was diammonium phosphate. The nine fertilization treatments were as follows: no N and no P applied (CK), 120 kg ha -1 N applied and no P (N1P0), 240 kg ha −1 N applied and no P (N2P0), 45 kg ha -1 P applied and no N (N0P1), 90 kg ha -1 P applied and no N (N0P2), 120 kg ha -1 N and 45 kg ha -1 P applied (N1P1), 240 kg ha -1 N and 45 kg ha -1 P applied (N2P1), 120 kg ha -1 N and 90 kg ha -1 P applied (N1P2), 240 kg ha −1 N and 90 kg ha −1 P applied (N2P2). N fertilizer was applied in two parts; 25% as base fertilizer and 75% at the flowering period, while P fertilizer was applied entirely as base fertilizer. The soil on the Loess Plateau is rich in potassium (K), and the soil total K is relatively high, so that farmers generally apply fertilizers without potassium [23,24]. The millet was  Basic soil physical and chemical properties in the study site were measured at a depth of 0-40 cm before fertilization (Table 1). Soil bulk density was measured using a soil bulk sampler and soil pH was measured using a pH meter. Notes: SOM = soil organic matter; TN = soil total nitrogen; TP = soil total phosphorus; all in (g·kg −1 ).

Experimental Design and Fertilizer Management
Local heat and moisture conditions dictate that crops are only grown once a year. Millet was planted in 2017 and soybeans were planted in 2018. The millet cultivar was Hongyang 7. The soybean variety planted was Zhonghuang 35. The order of nutrient absorption in millet was N > K > P. There were differences in nutrient concentrations among the millet crop organs, with P and K concentrations being the highest in the fruit, and the N concentrations being the highest in the root system. From the flowering period, the demand for N fertilizer began to increase rapidly to support grain growth. Nitrogen fertilizer could be applied at this period for optimal yield benefits. For every 100 kg of millet produced, 2-5.23 kg N, 0.2-0.75 kg P (P 2 O 5 ), 1.0-1.98 kg K (K 2 O) are required. The order of nutrient absorption in soybean was N > K > P. Soybean stakes are more fattening crops, because the formation of nutrients such as protein and fat in soybeans requires high nutrient amounts, especially N, P, and K. For every 100 kg of soybean produced, 5.3-7.2 kg N, 1-1.8 kg P (P 2 O 5 ), and 1.3-4.0 kg K (K 2 O) are required.
The experimental site is a rectangular block with an area of 567 m 2 , with nine fertilizer treatments, each with three replicates, in 3 × 7 m plots ( Table 2 and Figures 2 and 3). The N fertilizer was urea, while the P fertilizer was diammonium phosphate. The nine fertilization treatments were as follows: no N and no P applied (CK), 120 kg ha −1 N applied and no P (N1P0), 240 kg ha −1 N applied and no P (N2P0), 45 kg ha −1 P applied and no N (N0P1), 90 kg ha −1 P applied and no N (N0P2), 120 kg ha −1 N and 45 kg ha −1 P applied (N1P1), 240 kg ha −1 N and 45 kg ha −1 P applied (N2P1), 120 kg ha −1 N and 90 kg ha −1 P applied (N1P2), 240 kg ha −1 N and 90 kg ha −1 P applied (N2P2). N fertilizer was applied in two parts; 25% as base fertilizer and 75% at the flowering period, while P fertilizer was applied entirely as base fertilizer. The soil on the Loess Plateau is rich in potassium (K), and the soil total K is relatively high, so that farmers generally apply fertilizers without potassium [23,24]. The millet was fertilized with urea on 27th of April 2017, and supplementary nitrogen fertilizer was added on 3rd of July. Urea fertilizer was added to soybeans on 28th of April 2018, and supplementary nitrogen fertilizer was added on 26th of July. fertilized with urea on 27th of April 2017, and supplementary nitrogen fertilizer was added on 3rd of July. Urea fertilizer was added to soybeans on 28th of April 2018, and supplementary nitrogen fertilizer was added on 26th of July.

Sampling and Measurement
Soil sample collection method: taking each plot as a unit, the soil sampling point was more than 1 m away from the boundary of the sample plot. Sampling points were randomly selected with an S shape. Surface debris was then removed and soil samples were collected. All soil samples from the same sampling point were combined and were then passed through a 2 mm sieve. Samples were stored and labeled after the relevant sampling information had been recorded. In all the plots, soil samples were collected at 10 cm intervals up to a depth of 100 mm. Soil moisture was measured every 10 cm with a neutron moisture meter up to a depth of 100 cm. The soil samples were transported to the laboratory and were air-dried naturally, plant roots and other impurities removed, and the SOM, soil total N, and soil total P measured. The water consumed by weight was multiplied with the soil bulk density to obtain the water consumption by volume. Crop yield in the plots was expressed per m 2 , based on three 1 m 2 quadrats randomly selected in each plot. Furthermore, soil water consumption was determined using the oven-drying method. SOM, total N, and total P were determined using the H2SO4-K2Cr2O7 method [25], Kjeldahl method [26], and colorimetrically using the ammonium molybdate method [27], respectively. The 2017 millet was harvested on 12th of October, and the 2018 soybean crop was harvested on 4th of October. In 2017, soil moisture content before sowing was measured on 26th of April, and post-harvest moisture was measured on 13th of October. In 2018, soil moisture content before sowing was measured on 20th of April, and postharvest moisture was measured on 5th of October.

Calculations and Statistics
The water stored in the soil, water consumption during the crop growing season, and WUE were calculated using the following equations (Equations (1)-(3)):

Sw = H × D × B × 10
(1) where Sw (mm) is the sum of soil water stored in different soil layers; Di (g cm −3 ) is the soil bulk density in different soil layers; Hi (cm) is the soil depth; B is the percentage of soil moisture in weight; ET (mm) is the water consumption during the crop-growing season. W1 (mm) is the water storage at 0-

Sampling and Measurement
Soil sample collection method: taking each plot as a unit, the soil sampling point was more than 1 m away from the boundary of the sample plot. Sampling points were randomly selected with an S shape. Surface debris was then removed and soil samples were collected. All soil samples from the same sampling point were combined and were then passed through a 2 mm sieve. Samples were stored and labeled after the relevant sampling information had been recorded. In all the plots, soil samples were collected at 10 cm intervals up to a depth of 100 mm. Soil moisture was measured every 10 cm with a neutron moisture meter up to a depth of 100 cm. The soil samples were transported to the laboratory and were air-dried naturally, plant roots and other impurities removed, and the SOM, soil total N, and soil total P measured. The water consumed by weight was multiplied with the soil bulk density to obtain the water consumption by volume. Crop yield in the plots was expressed per m 2 , based on three 1 m 2 quadrats randomly selected in each plot. Furthermore, soil water consumption was determined using the oven-drying method. SOM, total N, and total P were determined using the H 2 SO 4 -K 2 Cr 2 O 7 method [25], Kjeldahl method [26], and colorimetrically using the ammonium molybdate method [27], respectively. The 2017 millet was harvested on 12th of October, and the 2018 soybean crop was harvested on 4th of October. In 2017, soil moisture content before sowing was measured on 26th of April, and post-harvest moisture was measured on 13th of October. In 2018, soil moisture content before sowing was measured on 20th of April, and post-harvest moisture was measured on 5th of October.

Calculations and Statistics
The water stored in the soil, water consumption during the crop growing season, and WUE were calculated using the following equations (Equations (1)-(3)): where S w (mm) is the sum of soil water stored in different soil layers; D i (g cm −3 ) is the soil bulk density in different soil layers; H i (cm) is the soil depth; B is the percentage of soil moisture in weight; ET (mm) is the water consumption during the crop-growing season. W 1 (mm) is the water storage at 0-100 cm soil depth during sowing. W 2 (mm) is the water stored at 0-100 cm soil depth in the harvesting season; P (mm) is the precipitation during the overall crop growth period; WUE represents water use efficiency (kg ha −1 mm −1 ); and Y is the crop yield (kg ha −1 ).No irrigation was carried out during any experimental year. The area where the experiment was carried out was flat. Therefore, the present study ignored soil water run-off and deep drainage.
One-way analysis of variance was used to examine the effects of fertilization treatment on SOM, soil total N, soil total P, soil water consumption, soil WUE, and crop yield (p < 0.05). Before the statistical analyses, we performed tests of normality and homogeneity. A Pearson's correlation analysis was also performed to explore the relationships among SOM, soil total N, soil total P, soil water consumption, soil WUE, and crop yield. All statistical analyses were performed using IBM SPSS Statistics ver. 26.0 (IBM Corp., Armonk, NY, USA). The differences between the treatments were calculated using the least significant difference test at 0.05 probability level. Figures were illustrated using Origin ver. 9.0 (OriginLab, Northampton, MA, USA).

Crop Yield
When

Correlation among Different Parameters
In 2017, the soil water consumption, soil WUE and yield were significantly positively correlated with SOM and soil total N (Table 3). In 2018, only soil water consumption was significantly positively correlated with SOM (Table 3). Soil WUE was significantly positively correlated with yield in the two experimental years (Table 3).

Discussion
Previous studies have demonstrated that fertilization increases nutrient concentrations in cropland soils [28,29]. In the present study, following nine fertilization experiments, we observed that SOM and total N content increased significantly when compared to the control conditions, by 27.1-81.3%, and 301.3-669.2%, respectively, only under combined N and P application, over a two-year period.
Fertilization significantly increased crop yield, hypothesis 1 was confirmed; while combined N and P fertilization (240 kg ha −1 N and 90 kg ha −1 N) resulted in the greatest increase in crop yield, hypothesis 2 was confirmed, after the single P fertilizer application, and single N fertilizer application. The control treatment, which consisted of neither P nor N fertilizer application, had the least effect on crop yield (Figure 7). N and P are essential nutrients for crop growth and development. The absorption and assimilation of N and P influence crop growth and development, and in turn, crop yield [30][31][32].
Studies have also demonstrated that soil bulk density decreases following fertilizer application, resulting in improved effective moisture diffusivity, soil thermal conductivity, and soil gas ratio, which promote crop growth and development [33]. Increased water retention following N and P fertilizer application is conducive to the formation, transfer, and absorption of soil nutrients, and promotes the decomposition of fertilizer, improves fertilizer utilization, reduces soil compaction, improves the tightness of soil particles, and facilitates adequate soil oxygen supply, in addition to reducing root penetration resistance [34]. A previous study reported that N and P fertilizer application in combination did not result in increased crop yields [28], which is inconsistent with the findings of the present study. The inconsistent observations could be due to different climatic conditions, soil texture, and rotation systems in the two studies.
Our results also demonstrate that fertilization can increase soil water consumption and soil WUE, and hypothesis 1 was confirmed. These increased from 1.6-7.8% and 79.5-515.2%, respectively, when compared to CK, following fertilization. Fertilization could increase soil nutrients and increase evapotranspiration rates in the leaves, in turn, improving the upward translocation of deep soil water for absorption and utilization by crops, which would improve soil moisture bioavailability [35]. Fertilization could also increase soil total porosity, total water consumption by crops, and soil WUE [36]. Numerous studies have demonstrated that fertilization practices influence soil water consumption significantly. The application of high rates of N and P fertilizer jointly led to the consumption of higher soil water amounts due to higher crop yield [37], indicating that higher rates of fertilizer application increase soil water consumption. In other studies, high fertilization rates improved soil properties and facilitated root extension to deeper levels and enhanced nutrient supply [38]. Increased root proliferation increases the volume of soil colonized, thereby reducing the probability of plant growth limitation during intermittent periods of drought [39,40].
We observed that using the N-P compound fertilizer (240 kg ha −1 nitrogen and 90 kg ha −1 phosphorus) in the present study, hypothesis 2 was confirmed, making it an appropriate fertilization treatment for improving soil water consumption and soil WUE. This treatment could facilitate the effective exploitation of the limited water resources in dryland regions where efforts are being made to implement sustainable cropping systems for dryland crops. The combined N and P application could promote the formation of a favorable soil aggregate structure (increasing the proportion of water-stable aggregates >0.25 mm in the soil) and improve soil microbial activity, which would, in turn, accelerate fertilizer use, and enhance the capacity of soil to maintain fertilizer supply [41]. Combined N and P fertilization would provide the N required for the growth of vegetative and reproductive organs, thereby increasing yield [42,43]. Combined N and P fertilization could maintain stable pH levels in soil, decrease the rate of soil acidification, improve the rhizosphere environment, and enhance soil capacity to maintain fertilizer supply to crops [44,45]. Furthermore, combined N and P fertilizer application could improve soil pore water and improve the soil micro-ecosystem, which would enhance soil conservation and the capacity to maintain fertilizer supply.
We also observed that soil water consumption, soil WUE, and crop yield were positively correlated with SOM and soil total N in 2017. The results indicated that the soil WUE and crop yield increased with an increase in soil nutrients. However, in 2018, soil WUE and crop yield did not exhibit any correlation with soil nutrients concentration, which could be attributed to changes in the physiological conditions of the planted crops. Soybeans are leguminous plants and have a high N-fixing capacity, so that they are less affected by N nutrient concentrations.

Conclusions
Rational fertilizer application could facilitate sustainable and effective exploitation of available rainfall resources. The results of the present study revealed a single application of treatment fertilizer and combined application of N and P fertilizers could increase crop yield and soil water consumption, and improve soil WUE, while satisfying crop growth and development requirements. In addition, combined N and P application (240 kg ha −1 N and 90 kg ha −1 P) increased soil WUE and crop yield in dryland agricultural production to a greater extent, suggesting that the combined application of N and P fertilizer (240 kg ha −1 N and 90 kg ha −1 P) should be promoted in arid and semi-arid areas to enhance agricultural production.
Author Contributions: The research design was completed by Q.L. and H.X. The manuscript was written by Q.L. and X.M. The collection and analysis of samples were performed by G.Z., P.G. and W.S. All authors have read and agreed to the published version of the manuscript.