Successive field experiments were conducted from 2013 to 2018 at the experimental station of the Heyang Dryland Agricultural Research Station of Northwest A & F University, located in the Heyang County of Shaanxi Province (35°19’ N, 110°4’ E, and 877 m above sea level), in the southeast region of the Loess Plateau in northwest China. At the experimental site, the average annual precipitation is approximately 494 mm (2004–2017), with approximately 60% of the annual rainfall occurring in July–September. The soil type is dark loessial soil and is classified as middle loam soil, according to the FAO/UNESCO Soil Classification (1993).

The experiment was arranged in a split-plot design with three replications. Planting density was assigned to the main plots, and maize cultivar was assigned to subplots. Four planting densities were evaluated in the experiment as follows: 52,500 plants ha^–1 (D1), 67,500 plants ha^–1 (D2), 82,500 plants ha^–1 (D3), and 97,500 plants ha^–1 (D4), with a row-to-row spacing of 50 cm. Three cultivars with different levels of tolerance to planting density were used in the experiment as follows: Yuyu22 (C1), Zhengdan958 (C2), and Xianyu335 (C3) (Xue et al., 2010). Other field management practices followed in this study have been described previously (Zhang et al., 2019c).

Daily weather datasets (i.e., solar radiation, air temperature, and rainfall) were obtained from the national meteorological database,¹ and the data from 2013 to 2018 are shown in Figure 1.

Five plants were randomly selected at different stages to determine the green leaf area (leaf length × leaf width × 0.75) and LAI (total leaf area per ha) of each maize plant (Zhang et al., 2019b). After measuring leaf area, the same plants were used for measuring the aboveground biomass. To measure the aboveground biomass, plants were fixated at 105°C for 0.5 h and then oven-dried at 85°C for a minimum of 48 h until a constant weight was achieved.

Five plants were randomly selected from each plot at the jointing (V6), tasseling (VT), and filling (R3) stages, and the net photosynthetic rate (Pn), transpiration rate (Tr), and stomatal conductance (Gc) of leaves were measured using a Li-6400 portable photosynthesis system (Li-COR Inc., Lincoln, NE, United States). These measurements were taken between 9:00 a.m. and 11:00 a.m. on a clear sunny day. The largest leaf was sampled at the V6 stage, while the maize ear leaf was sampled at the VT and R3 stages. Leaf chlorophyll content was determined using photometric methods, as described by Cui et al. (2019).

The IPAR (MJ m^–2) per plant canopy and radiation use efficiency (RUE) (g MJ^–1) data were determined using the following equations (Zhang et al., 2019b):

where R is the daily solar radiation (MJ m^–2 day^–1), k is the light extinction coefficient (0.65 for maize), and LAI is the LAI.

In each treatment, three random quadrats covering a 9.0 m² area were selected to determine yield and yield components (kernel number per square meter and 100-kernel weight). Grain and biomass yield were determined at 14% moisture content. HI and precipitation use efficiency (PUE) were calculated using the following equations:

where P is the amount of precipitation (mm) during the growing season.

Crop yield stability, as affected by different treatments, was evaluated based on its variability by measuring the coefficient of variation (CV, %) using the following equation (Xu et al., 2019):

where STD(Yt) is the SD of grain yield of a particular treatment over the 6-year experiment period, and AVE(Yt) is the mean yield of that treatment over the same period.

The sustainable yield index (SYI) is a quantitative measure to assess the sustainability of any agricultural system (Sharma et al., 2013). The SYI was calculated using the following equation (Li et al., 2016):

where Ymax represents the maximum crop yield attained by any treatment during the study period, and AVE(Yt) is the mean yield of that treatment over the same period.

The sampled maize plants were separated into different organs. Samples were then oven-dried at 85°C to measure the dry matter weight. Nitrogen concentration in plant samples was analyzed based on the Kjeldahl method (Du et al., 2021). Nitrogen uptake, nitrogen harvest index (NHI), nitrogen use efficiency (NUE), nitrogen productive efficiency (NPE), and nitrogen uptake efficiency (NUPE) were calculated as follows (Zhang et al., 2019b):

The statistical significance of density, cultivar, and their interaction was assessed with two-way ANOVA. All data were analyzed using the IBM SPSS statistical software package (version 20.0, SPSS Inc., Chicago, IL, United States), followed by the least significant difference (LSD) test. Differences among treatments were considered statistically significant at p < 0.05, and figures were generated using Origin 2015 (v. Pro 2019; OriginLab Corp., Northampton, MA, United States).

Maize biomass yield varied significantly with planting density and cultivar over the six cropping seasons (p < 0.05) (Table 1). Aboveground biomass accumulation increased with the increase in planting density (Figure 2), with the highest value recorded in the D4 treatment. Biomass yield accumulation increased slowly from the V3 to V6 stage and rapidly from the V6 to VT stage, with the highest value recorded at physiological maturity (Figure 2). In contrast, HI decreased with the increase in planting density (Table 1).

Yield and its components were significantly affected by density and cultivar over the 6 years (p < 0.05) (Table 1). The average ear number per square meter increased, with the increase in planting density, whereas the 100-kernel weight decreased. Grain yield did not increase with the increase in planting density and showed the highest value in the D2 treatment (Table 1). The interaction between density and cultivar had significant effects on yield and its components (p < 0.05). The yield variation (CV) increased with the increase in planting density, but the SYI value decreased (Table 2). Differences in yield stability were detected among the three cultivars, and the C2 and C3 showed lower yield variation than the C1. These results indicate that high planting density raises the yield variability and decreases the yield sustainability of dryland maize, which was not conducive to the sustainable development of dryland farming.

Previous research has demonstrated that increasing planting density improves maize canopy closure, i.e., rapid canopy establishment and leaf area expansion, leading to greater IPAR, which contributes to greater radiation capture (Teixeira et al., 2014; Du et al., 2021). Similar results were obtained in this study. Compared with the D1 treatment, the average LAI values of the three cultivars increased by 19–27%, 38–44%, and 45–60% in D2, D3, and D4 treatments, respectively. After the VT stage, the LAI value decreased due to the shedding and senescence of plant leaves, but the amplitude of this decline was small in the low density, which is beneficial to the assimilating of photosynthetic products and resulting in higher partitioning of carbohydrates to the ear. This was mainly related to the lower interplant competition between plants, which has been reported in maize (Hammer et al., 2009; Rossini et al., 2011). Additionally, low-density crops maintain high green leaf area and leaf chlorophyll content and were accompanied by higher photosynthetic characteristics, such as Pn and Gc (Figure 4). The Gc affects the exchange of CO₂ and H₂O between leaves and the environment, as an adaptive mechanism to cope with drought stress (Hernández et al., 2020). Zhu et al. (2010) showed that photosynthetic efficiency is closely related to the regulation of stomatal opening and leaf chlorophyll content, and the increase in crop productivity relies on improved photosynthesis. Thus, optimizing canopy structure and maintaining photosynthetic capacity while increasing the resource use efficiency would be the key to improve the maize yield by optimizing planting density. One limitation of this study is that we monitored the photosynthetic characteristics and chlorophyll content of only the ear leaves, and the photosynthetic performance of the whole maize population remains unknown. Further investigation will help explain yield formation from the perspective of group light energy efficiency.

In this study, biomass yield increased with the increase in plant density, whereas grain yield showed a parabolic relation with planting density (Table 1). The increasing of planting density results in lower light intensity in the canopy, but a certain grain yield needs more leaf area (to realize a high canopy photosynthesis rate) to support its grain filling and crop yield (Du et al., 2021). Thus, the HI decreased dramatically with increasing planting density. Increasing planting density significantly improved LAI and IPAR of the canopy, eventually resulting in a significant increase in aboveground dry matter accumulation (Teixeira et al., 2014). However, as planting density increased, the photosynthetic characteristics of plants declined, resulting in lower crop photosynthetic assimilation and productivity per plant, which might explain the decrease in maize yield observed in this study at high planting density. These results indicate that dryland maize productivity at high planting density is limited by the relatively low translocation of assimilates to grains. Therefore, pursuing high planting density is not a desirable strategy in the rainfed farming system, while the relatively lower planting density may be more conducive to the effective use of limited resources of semiarid environments. Increasing planting density also increased the yield variability (CV, %) and decreased the yield sustainability of dryland maize (Table 2). Mylonas et al. (2020) also revealed that CV (%) values of plant yield increased when planting density increased, mainly due to increased competition for resources, especially for soil water in rainfall agroecosystems. While under lower planting density, the available water per plant increases, which can maintain the growth of crops and filling of grain. The cultivar is another factor affecting grain yield response to density and stability, as shown by previous studies (Berzsenyi and Tokatlidis, 2012; Chen et al., 2017; Solomon et al., 2017), as well as our current results. In this study, C2 and C3 showed higher yield and yield stability over the six cropping seasons than the C1 (Table 2). The lower yield of the C1 was associated with the rapid decline in LAI after the tasseling stage (Figure 3). This is consistent with previous findings reported that the reduction in green LAI results decreases the fraction of total radiation intercepted and leads to lower carbohydrate remobilization from leaves to the ear (Xue et al., 2010).

Improving the resource use efficiency of crop plants is the main strategy to realize the sustainable development of agriculture. In this study, PUE was significantly affected by planting density and cultivar (p < 0.05) (Table 1). A similar trend was shown by grain yield, and PUE did not increase with the increase in planting density but showed a parabolic relation with planting density. Results by Tokatlidis et al. (2011) and Berzsenyi and Tokatlidis (2012) highlighted the importance of maize cultivars that are less dependent on high planting density to increase resource use efficiency in non-irrigated land. Although increments in IPAR were in accordance with increasing LAI, they did not promote higher RUE. This is partly because light attenuation within the canopy was increased under higher plant population due to shading, and relatively more light captured by the upper canopy has been suggested to reduce the whole plant photosynthetic efficiency, which in turn decreases the RUE (Du et al., 2021). In addition to water and radiation, crop productivity also depends on the absorption of nutrients and allocation of assimilates (Teixeira et al., 2014; Xu et al., 2017; Zhang et al., 2019a). In this study, increasing plant population did not increase the NUPE and NUE over 6 years (Table 4). This was mainly because increasing planting density decreases the capacity of the crop to accumulate nitrogen per unit green LAI (Ciampitti and Vyn, 2011), thus decreasing the NPE. Therefore, provided cultivars have high plant yield efficiency, and using lower planting density to enhance crop resilience to extremely fluctuating environments will be more meaningful for the long-term development of dryland agriculture. Differences in cultivar characteristics are one of the main reasons for the differences in resource use efficiency (i.e., radiation, water, and nutrients). Density-tolerant cultivars (C2 and C3) exhibited higher resource use efficiency than the C1 under the same climatic conditions. This was mainly related to the light distribution through the canopy, which was increased for density-tolerant cultivars due to their upright leaves and small leaf angles (Xue et al., 2010). This resulted in relatively more light being captured by the lower canopy of density-tolerant cultivars, thus improving their resource use efficiency.