Seeding Rate Effects on Hybrid Spring Wheat Yield, Yield Components, and Quality

: Agronomic practices, such as planting date, seeding rate, and genotype, commonly inﬂuence hard red spring wheat (HRSW, Triticum aestivum L. emend. Thell.) production. Determining the agronomic optimum seeding rate (AOSR) of newly developed hybrids is needed as they respond to seeding rates differently from inbred cultivars. The objectives of this research were to determine the AOSR of new HRSW hybrids, how seeding rate alters their various yield components, and whether hybrids offer increased end-use quality, compared to conventional cultivars. The performance of two cultivars (inbreds) and ﬁve hybrids was evaluated in nine North Dakota environments at ﬁve seeding rates in 2019 − 2020. Responses to seeding rate for yield and protein yield differed among the genotypes. The AOSR ranged from 3.60 to 5.19 million seeds ha − 1 and 2.22 to 3.89 million seeds ha − 1 for yield and protein yield, respectively. The average AOSR for yield for the hybrids was similar to that of conventional cultivars. However, the maximum protein yield of the hybrids was achieved at 0.50 million seeds ha − 1 less than that of the cultivars tested. The yield component that explained the greatest proportion of differences in yield as seeding rates varied was kernels spike − 1 (r = 0.17 to 0.43). The end-use quality of the hybrids tested was not superior to that of the conventional cultivars, indicating that yield will likely be the determinant of the economic feasibility of any future released hybrids.


Introduction
Current global crop production needs to double to meet the projected global demand by the year 2050 with wheat yield gains needing to increase from 0.9 to 2.4% yr −1 [1]. Maximizing crop production efficiency and profitability is challenged by market instability [2] and the increasing production costs are often a critical barrier to achieving these goals. Yield is the primary factor affecting profitability when paired with the efficient use of crop production inputs, such as fertilizers and seed [3]. Hard red spring wheat (HRSW) grain yield is determined by the complex interaction between genotype, environment, and management [4]. Environment cannot be fully controlled nor accurately predicted, however, producers can select and anticipate the effect of management practices, such as crop rotation [5], genotype [6], seeding date [7], seeding rate [5,7,8], row spacing [8], weather and soil management [9], genetic disease control [10], and pest management [11].
Since the advent of hybrid maize (Zea mays L.) [12], the exploitation of heterosis as a means of increasing yields in various crops, such as rice (Oryza sativa L.) and barley (Hordeum vulgare L.) has been explored [13][14][15]. Interest in exploiting wheat's heterosis began after male-sterility advances were reported by Kihara [16], Fukasawa [17], and Kihara [18]. Livers and Heyne [19] found hybridized wheat genotypes yielding 30% more than the best performing inbred cultivar at the time. More recent research has shown a 20% yield improvement in hybrids in comparison to the best commercial cultivar and greater yield stability between environments [20]. European winter wheat hybrids displayed a higher nitrogen-use efficiency, compared to inbred cultivars [21]. The adoption of hybrid Yield components are usually reported as a representation of the final makeup of yield. These components do not develop independently and there is a complex relationship between the various yield components [42][43][44][45]. Some components, such as tiller numbers, are more plastic than others, such as kernels per spike [46]. However, environment and genetics strongly influence yield component composition [45,46].
The objective of this research was to determine the AOSR for new spring wheat hybrids, determine the contribution of their various yield components to yield, and to determine if hybrids have increased end-use qualities relative to conventional cultivars.

Materials and Methods
Field experiments were established in three environments in 2019 and two environments in 2020. In 2019, experiments were located in Hettinger, Langdon, and Minot, North Dakota, USA, representing a large geographical area of HRSW production. In 2020, field experiments were conducted in Grand Forks and Prosper, North Dakota, USA. Table 1 summarizes the physical characteristics of the experimental locations. Treatments consisted of a factorial combination of seven genotypes (two inbred cultivars and five HRSW hybrids) and seeding rates in a randomized complete block design with a split-plot arrangement with four replications. The whole plot was seeding rate and the sub-plot was genotype. In most environments, there were two planting dates, the first date, which was considered the optimal date, occurred as soon as practical in the spring as recommended by Wiersma and Ransom [47]. The second date was two weeks thereafter. Seeding rates were 2.22, 2.96, 3.71, 4.45, and 5.19 million live seed ha −1 which will be referred to using the million seeds ha −1 unit. The genotypes included were "SY Ingmar" and "SY Valda" (inbred cultivars), which will be referred to as Ingmar and Valda, and five experimental hybrids which will be identified as H A , H B , H C , H D , and H E . All genotypes were developed by AgriPro (Syngenta AgriPro, LLC, Greensboro, NC, USA). Ingmar and Valda were among the most commonly grown cultivars and occupied 33% of the total HRSW hectares cultivated in North Dakota in 2019 [48]. Both the agronomic traits and disease resistance of these cultivars are summarized in (Table 2). The hybrids included had not yet been commercially released. Their pedigrees were not disclosed and, because of no previous testing, their agronomic characteristics were not known.
The quantity of seed planted for each genotype and seeding rate was calculated on a live seed basis based on a germination test. Plot size, seeding date, and harvest date information are summarized for each location in Table 3. Management of the trials varied slightly at each location, based on the preferred practices of cooperating researchers at the Hettinger, Langdon, and Minot locations. Soils were tested for plant essential nutrients before seeding to ensure fertility was not a limiting factor (Table 4) and N, P, and K rates were adjusted to ensure they were not limiting as recommended by Wiersma and Ransom [47]. Fertilizers were applied before planting if necessary.  3 Straw strength = 1 to 9 scale, with 1 the strongest and 9 the weakest [48]. 4 Disease reaction scores from 1 to 9, with 1 = resistant and 9 = very susceptible [48].  The fungicide combination of pydiflumetoen (150 g ai ha −1 ) and propiconazole (126 g ai ha −1 ) commercially marketed as Miravis Ace (Syngenta Crop Protection, LLC, Greeensboro, NC, USA) was applied to all locations at Zadoks 65 at a rate of 0.164 L ha −1 to reduce Fusarium head blight incidence (Fusarium graminearum) and fungal leaf spots. Other management practices were based on North Dakota State University Extension recommendations, regarding cultivation, fertilization, and herbicide and pesticide applications [47].
Plant density and spike density were obtained by counting plants and spikes in two of the innermost rows of each plot from a 0.91 m length from a stake randomly placed after sowing. Plant density was determined at approximately Zadoks 11 and prior to tiller production. Productive spike density was determined by counting spikes at approximately Zadoks 92 within the same 0.91 m of rows used for plant density measurements. Small spikes that were deemed not to contribute to yield were not counted.
Spikes plant −1 and kernels spike −1 were derived from plant density, spike count, and kernel weight measurements. Yield was collected for each plot using a small plot combine and was adjusted to 13.5% moisture. Moisture and test weight were determined using a GAC 2100 moisture tester (Dickey-John Corp., Minneapolis, MN, USA). Percent grain protein content was measured using a DA 7250 NIR analyzer (Perten Instruments, Stockholm, Sweden) and was reported on a 12% moisture basis. The moisture contents used for reporting yield and protein content were those commonly used by the wheat industry in the USA. To report protein yield, protein and yield, which are commonly reported at 12 and 13.5% moisture, respectively, were corrected to 12% moisture and multiplied.
An ANOVA was performed on each measured variable using the MIXED procedure in SAS 9.4 (SAS Institute, Cary, NC, USA). Environments with the ratio of the highest and lowest error mean square for yield, protein, or protein yield being less than 10 were combined for analysis [49]. In the combined analysis, location-year were considered an environment and a random effect while seeding rate, genotype, and the seeding rate by genotype interaction were considered fixed effects and analyzed as described by Carmer [50]. Analysis of variance was conducted for yield (seven environments), protein (eight environments), and protein yield (eight environments). Significant ANOVA seeding rate by genotype interactions for yield and protein yield were analyzed using the MIXED procedure with treatment means separated by Tukey's multiple comparison test at P = 0.05. The seeding rate by genotype interaction term was further analyzed using regression analysis and the REG procedure in SAS.
The wheat quality analysis was performed on the grain of each genotype from the 2.96 million seeds ha −1 seeding rate treatment. The 2.96 million seeds ha −1 rate was chosen to represent a typical seeding rate that farmers use in the area. The harvested samples from replicates one and two and from replicates three and four were combined for each of these treatments from each location to form two replicates for each genotype and environment. Grain samples were analyzed by the North Dakota State University Wheat Quality Laboratory for test weight, kernel size distribution, kernel weight, protein, falling number, milling extraction, mixograph score, peak maximum time, maximum torque, total energy, and loaf volumes, according to standard protocols [51].
Planting dates within environments had differing yield responses, so data from each planting date were partitioned into high (>5000 kg ha −1 ) and low (<5000 kg ha −1 ) yielding environment datasets by considering individual planting dates as a single environment for a total of nine environments similar to Mehring et al. [52]. Yield within an environment was evaluated by standardizing the distribution of each, using z-scores. Data were transformed using the Standard procedure in SAS to calculate a z-score for yield using the formula z − score =(x − x)/σ 1 where x is yield, x is the yield mean of the high or low yielding environment, and σ 1 is the standard deviation of the high or low yield environment [53]. The z-score adjusts the data distribution to have a mean of 0 and a standard deviation of 1. Next, the z-score data were used to adjust yield values to be relative to the yield of the individual environment using the formulaŶ = (z-score × σ 2 ) + µ whereŶ is estimated yield, z-score is z-score of yield, σ 2 is the standard deviation of the individual environment, and µ is the mean of the high or low yield environment.
Relationships between yield components and yield for high and low yielding environments were analyzed using yields adjusted by the z-score approach previously discussed. Multiple linear regression was used to determine the relative importance of each yield component on yield using the REG procedure which also derived partial correlation coefficients (r 2 ) and adjusted R 2 . Wheat end quality characteristics were analyzed using the GLM procedure in SAS with single degree of freedom linear contrasts for cultivar and hybrid comparison evaluated at α = 0.05.

Weather Data
Seeding dates were dependent on appropriate seeding conditions which were primarily affected by rainfall each year (Figures 1 and 2). Seeding dates ranged from 116 to 149 day of the year (DOY) with recommended optimal seeding dates ranging from 91 to 127 DOY from southernmost to the northernmost latitudes in North Dakota. The yield variation between environments can be largely attributed to weather and soil effects, since the management factors were similar in 2019. Hettinger had relatively normal temperatures and rainfall during the growing season and yields were slightly above the mean for the region. In 2019, rainfall was less than normal in Minot and Langdon with Minot generally Agronomy 2021, 11, 1240 6 of 14 having warmer temperatures than Langdon. In Minot, minimal rainfall and above normal temperatures between 140 to 170 DOY during tiller formation (Zadoks 11 and 29) greatly reduced yield potential. The 2020 growing season began with above-normal rainfall and cooler than normal temperatures in Grand Forks and Prosper (Figure 2). The low early season temperatures and normal rainfall allowed for favorable tiller formation and grain development in Prosper, resulting in high grain yield. marily affected by rainfall each year (Figures 1 and 2). Seeding dates ranged from 116 to 149 day of the year (DOY) with recommended optimal seeding dates ranging from 91 to 127 DOY from southernmost to the northernmost latitudes in North Dakota. The yield variation between environments can be largely attributed to weather and soil effects, since the management factors were similar in 2019. Hettinger had relatively normal temperatures and rainfall during the growing season and yields were slightly above the mean for the region. In 2019, rainfall was less than normal in Minot and Langdon with Minot generally having warmer temperatures than Langdon. In Minot, minimal rainfall and above normal temperatures between 140 to 170 DOY during tiller formation (Zadoks 11 and 29) greatly reduced yield potential. The 2020 growing season began with above-normal rainfall and cooler than normal temperatures in Grand Forks and Prosper (Figure 2). The low early season temperatures and normal rainfall allowed for favorable tiller formation and grain development in Prosper, resulting in high grain yield.

Yield
There was a significant seeding rate by genotype interaction for yield when the data were combined across the environments (Table 5). However, the seeding rate by genotype interaction differences in Table 6 inconclusively described the genotype response to seeding rate. To further evaluate this interaction, the response of each genotype to increasing seeding rates was evaluated using regression analysis (Table 7). Significant regression R 2 values were relatively similar between linear or quadratic models for each genotype. The AOSR ranged from 3.60 to 5.19 million seeds ha −1 with an average AOSR for the hybrids

Yield
There was a significant seeding rate by genotype interaction for yield when the data were combined across the environments (Table 5). However, the seeding rate by genotype interaction differences in Table 6 inconclusively described the genotype response to seeding rate. To further evaluate this interaction, the response of each genotype to increasing seeding rates was evaluated using regression analysis (Table 7). Significant regression R 2 values were relatively similar between linear or quadratic models for each genotype. The AOSR ranged from 3.60 to 5.19 million seeds ha −1 with an average AOSR for the hybrids of 4.49 million seeds ha −1 which was similar to the average AOSR for the cultivars. However, H E had an AOSR that was 83% of the Valda yield, suggesting that highly productive hybrids may require a lower seeding rate than inbred cultivars, at least for those included in this trial. Our AOSR results are within the ranges of those for HRSW and winter wheat cultivars of 1.59 to 5.03 and 2.85 to 5.73 and million seeds ha −1 in the northern great plains and Ohio, USA, respectively [25,30]. In Italy, winter wheat hybrids can be planted at 33% of the seeding rate of a conventional cultivar while not losing significant amounts of yield [54]. In Europe, winter wheat hybrids display greater yield stability compared to conventional cultivars [55]. Similarly, rice hybrids have greater yield stability and are less responsive to reduced inputs [56]. Thus, certain HRSW hybrids can have lower AOSR and require fewer inputs compared to conventional cultivars; however, the hybrids and inbred cultivars in this experiment were generally unresponsive to seeding rate.

Grain Protein Content
Hard red spring wheat is often sought after for its high protein content for making raised bread products. Various studies have affirmed seeding rate often does not significantly affect grain protein [57][58][59]. Our results show that seeding rate did not affect grain protein content (Table 5). Conversely, Chen et al. [8] found a linear relationship where protein content decreased as seeding rate increased from 1.08 to 4.30 million seeds ha −1 . Genotypes differed significantly in protein content with Ingmar having the greatest protein content also being the lowest yielding among all genotypes (Table 8). Protein content levels were relatively consistent across the hybrids that were included in this experiment although significantly less than Ingmar. Thorwarth et al. [60] found winter wheat hybrids to have decreased protein content regardless of quality class, compared to parental lines. The data suggest hybrid wheat protein contents can be similar to relatively lower protein cultivars like Valda that have similar yield potential. Within columns, means followed by the same letter are not significantly different according to Tukey's multiple comparison test with p = 0.05. 1 Yield, protein, and protein yield were determined at 13.5%, 12%, and 12% moisture content using seven, eight, and eight environments, respectively.

Protein Yield
The protein yield accounts for the grain protein percentage and seed yield to describe the amount of protein produced per area [61]. The relationship is usually negative between protein and yield, and the variable is useful for integrating them both, as protein content is critical in determining HRSW value. The protein yield was significantly affected by seeding rate, genotype, and the seeding rate by genotype interaction ( Table 5). The significant seeding rate by genotype interaction in Table 6 is difficult to interpret, therefore regression analysis was used. Using regression analysis of the protein yield, Valda and H A were similar with H A yielding 18 kg ha −1 more protein (Table 9). Across the five hybrids which ranged from 615 to 634 kg ha −1 protein yield, only H A had a significant protein yield response to seeding rate (Table 9). In general, neither linear nor quadratic relationships between protein yield and seeding rate explained the relatively high amount of deviation from the trendline, aside from Valda and H A . Certain HRSW genotypes may elicit a response in protein yield whereas others are unresponsive. On average, hybrids had a greater protein yield compared to their inbred parental lines [60]. Based on these quadratic protein yield models, the HRSW hybrids do not have a reduced protein yield at the expense of grain yield, compared to cultivars. Simmonds [62] summarized that protein yield increases with grain yield but diminishes beyond an optimal point.

Yield Components
Wheat yield components are considered plastic and compensate for one another and their relative contributions to yield may vary in high or low yielding environments. Therefore, the data were separated into four high (>5000 kg ha −1 ) and five low (<5000 kg ha −1 ) yield environments, similar to Mehring et al. [52]. The environments ranged from 5202 to 5923 and 3846 to 4633 kg ha −1 for high and low yield environments, respectively. Of the genotypes tested, Valda and H D displayed similar yield, quality, and growth characteristics and were used to compare yield components. Plant density did not significantly influence the yield of either Valda or H D in either yield environment (Table 10). In general, spike density and kernels spike −1 contribute to the greatest proportion of yield. In high yielding environments, spike density and kernels spike −1 , and kernel weight are more closely related to yield for H D compared to Valda. The differences between the yield component contributions to yield for H D and Valda were not as apparent in low-yielding environments. Standardizing the data for multiple linear regression allows for the relative comparison of each yield component's contribution to yield. The multiple linear regression coefficients for Valda and H D in high yield environments show that the relative importance of the yield components are similar. Yield components in high yield environments better explain yield for H D (R 2 = 0.79) than Valda (R 2 = 0.56). Furthermore, yield component contributions to yield for Valda and H D have no discernable differences in low-yield environments. Table 10. Cultivar Valda and H D in four high (>5000 kg ha −1 ) and five low (<5000 kg ha −1 ) yield environments with partial correlation coefficients (r) and significance for relationships between the yield and yield components and the corresponding standardized multiple linear regression equations. Spike density, kernel weight, and kernels spike −1 consistently explained yield for both Valda and H D in both high and low yield environments. Our results show that in the hybrids, kernels spike −1 were increased, compared to the inbred cultivar Valda, confirming previous studies [21,63,64]. Slafer et al. [45] described kernels spike −1 as a coarse yield regulator accounting for large changes in yield caused by genotypic differences and can be targeted with management practices to improve overall yield. Improving wheat yield potential through maximizing spike density, kernels spike −1 , and kernel weight may benefit hybrid wheat more than inbred cultivars if used for selection during the breeding process. In particular, kernels spike −1 is a more important yield component for HRSW hybrids, compared to inbred cultivars.

Grain Quality
For each genotype, the seed from all replicates was combined over the 2.96 million seed ha −1 seeding rate for each environment (Table 11). The hybrids had significantly greater proportions of small and medium-sized kernels and greater kernel weight in general. The hybrids also had less milling extraction compared to Ingmar and Valda. This could be expected, given larger kernels are more desirable for milling and provide a higher flour yield compared to medium and small kernels [65]. However, when compared to Ingmar and Valda, the wheat hybrids did not have superior end-use qualities. The hybrids had a relatively similar protein content compared to Valda. In general, experimental wheat hybrids in breeding programs have negative heterosis for grain protein content compared to inbred cultivars [60]. Ingmar and Valda had greater falling numbers, although none of the genotypes had falling numbers that exceeded the preferable value of 400 s. Kindred et al. [66] similarly found hybrid wheat falling numbers to be less than the inbred parent cultivars. These results correspond with those of Gaines et al. [67] who found soft wheat milling extraction was less for smaller kernel sizes although other end quality characteristics were not affected by small kernel sizes.