Agronomy Journal Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 27 June 2006
Published in Agron J 98:1137-1145 (2006)
DOI: 10.2134/agronj2006.0039
© 2006 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Agricola
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Related Collections
Right arrow Other Crop Management
Right arrow Best Management Practices
Right arrow Wheat

Wheat

Minimizing Protein Variability in Soft Red Winter Wheat

Impact of Nitrogen Application Timing and Rate

Dianne C. Farrera, Randy Weisza,*, Ronnie Heinigera, J. Paul Murphya and Jeffrey G. Whiteb

a Dep. of Crop Science, North Carolina State Univ., Raleigh, NC 27695-7620
b Dep. of Soil Science, North Carolina State Univ., Raleigh, NC 27695-7619

* Corresponding author (randy_weisz{at}ncsu.edu)

Received for publication February 9, 2006.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Grain protein content in soft red winter wheat (Triticum aestivum L.) is highly variable across years and environments in the southeastern USA. This variability makes southeastern wheat undesirable to millers and negatively impacts its value in the export market. The objectives of this study were to determine how different N fertilizer rates and application times would affect grain protein variability and to determine if there were N fertilizer recommendations that would minimize regional protein variation. We conducted experiments in the North Carolina Piedmont, Coastal Plain, and Tidewater in 2001 and 2002. At each site–year, we used a split-plot design with three or five N fertilizer rates at growth-stage 25 (GS) (main plots), and an additional five N fertilizer rates applied at GS 30 (subplots). Analysis of variance indicated that environment contributed 68 and 90.5% of the variability in yield and test weight, respectively. Though environment contributed 23.3% of grain protein variability, the majority (51.4%) was attributed to timing and rate of N application. As grain protein levels increased at higher N rates, so did overall protein variability. Additionally, applying the majority of N fertilizer at GS 30 increased grain protein variability compared to application at GS 25. Based on these results, our recommendations to reduce grain protein variability in the southeastern USA are to: (i) reduce the range in N fertilizer rates used across the region, (ii) avoid overapplication of N beyond what is required to optimize yield and economic return, and (iii) apply spring N at GS 25.


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
THE demand for soft red winter wheat by the milling and baking industry in the southeastern USA continues to grow, and the region's millers generally pay a premium for locally grown high-quality grain. However, grain protein content in the southeastern USA is highly variable across years and environments. The milling and baking industry's desired range in grain protein content for soft wheat is 80 to 110 g kg–1 (Gooding and Davis, 1997). However, the desired grain protein ranges for specific uses such as cookies or pastry are narrower. Grain protein content in North Carolina has been reported to range from 86 to 135 g kg–1 across cultivars and environments (Bowman, 2002, 2003, 2004). This variability makes regional wheat undesirable to southeastern millers, who currently import approximately 50% of their soft red winter wheat from the midwestern USA, where grain protein content is generally more consistent. High protein variability in southeastern soft red winter wheat not only makes this grain less desirable to regional millers, but also lowers its value in the export market.

Variability in grain protein can be attributed to environments that differ across locations and years with respect to seasonal temperatures, moisture, and soil type (Gooding and Davis, 1997; Gooding et al., 1997). Variability in grain protein can also be attributed to differences in cultivar genetic potential and to management decisions (Miezan et al., 1977; Vaughan et al., 1990; Beuerlein et al., 1992; Smith and Gooding, 1996; López-Bellido et al., 1998). Among the most important management practices influencing grain protein content is N fertilizer application rate and timing. Increasing N fertilizer rates can result in higher grain protein content (Johnson et al., 1973; Nelson et al., 1978; Fowler et al., 1989; Vaughan et al., 1990; Kelley, 1995; Smith and Gooding, 1996; López-Bellido et al., 1998). Nitrogen fertilizer rate recommendations in the southeastern USA generally call for N to be applied at growth stage (GS) 25 and/or 30 (Zadoks et al., 1974), with total amounts at these two growth stages not to exceed 134 kg N ha–1 (Alley et al., 1994; Scharf et al., 1993; Scharf and Alley, 1993; Weisz et al., 2001; Weisz and Heiniger, 2004). However, it is not unusual for soft red winter wheat producers to apply spring N fertilizer rates that range from as low as 45 kg N ha–1 when the price of N is high or the crop appears to have low yield potential, to as high as 202 kg N ha–1. This wide range in N fertilizer rates may contribute to regional variability in wheat protein content.

Growers in many wheat production areas apply at least half of their total N fertilizer before planting. In the southeastern USA, however, warm temperatures and high precipitation generally result in denitrification and leaching of preplant N fertilizer (Counce et al., 1984; Scharf et al., 1993; Scharf and Alley, 1993). Therefore, the majority of N is typically applied at GS 30 if GS 25 tiller densities are high (above 550 tiller m–2), or in split applications (the first split at GS 25 and the second at GS 30) if GS 25 tiller density is low (Scharf et al., 1993; Scharf and Alley, 1993; Weisz et al., 2001). While this practice optimizes yield (Weisz et al., 2001), it may have an adverse impact on protein variability. In wheat production regions where the majority of N is applied at planting, delaying at least some of the N application until later growth stages has been shown to increase grain or flour protein content in some cases (Vaughan et al., 1990; Stark and Tindall, 1992; Kelley, 1995; Woodard and Bly, 1998). While we found no research on the effect of applying the majority of N fertilizer at GS 30 compared to splitting applications between GS 25 and GS 30, it seems likely that differences in how southeastern USA producers time their N applications might also contribute to regional grain protein variability.

Given the negative influence of grain protein variability on the marketability of soft red winter wheat in the southeastern USA, a determination of how N management influences grain protein variability is needed. In this light, our primary objectives were to determine how different N fertilizer rates and times of application would affect overall grain protein variability. Additionally, we wanted to compare the proportion of grain protein variability caused by different N treatments to that caused by environmental effects. Our secondary objective was to determine if there were N fertilizer recommendations that would minimize regional protein variation.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Experimental Environments
To encompass the range of soil and environmental variability representative of the region, experiments were conducted in the North Carolina Piedmont, Coastal Plain, and Tidewater in 2001–2002 and 2002–2003. Three site–years in 2001 and 2002 were located at the Piedmont Research Station near Salisbury, NC (P2001, P2001nt, and P2002; see Table 1). Site–years in the Coastal Plain region were the Cunningham Research Station located near Kinston, NC in 2001 and 2002 (C2001 and C2002), and the Lower Coastal Plain Tobacco Research Station located near Kinston, NC in 2001 (L2001). In 2002, an experiment was located at the Tidewater Research Station near Plymouth, NC (T2002). Taxonomic classification of the soils at these seven site–years is shown in Table 1.


View this table:
[in this window]
[in a new window]
  		Table 1. Site–year abbreviation (Site), location, soil series, and soil taxomomic classification for each experimental location in a study of the effects of N fertilizer timing and rate on wheat grain protein variability.

 
Experimental Design
In each site–year, a split-plot randomized complete block design with five replications was used. The main plot treatment consisted of N fertilizer rates applied at GS 25 ("N25"). At P2002, N25 rates were 0, 67.2, and 100.8 kg N ha–1. At all other site–years, five N25 rates (0, 33.6, 67.2, 100.8, and 134.4 kg N ha–1) were applied. The subplot treatments at all site–years consisted of these same five N rates applied at GS 30 ("N30"). The combination of main plots and subplots resulted in 25 N treatments consisting of different N fertilizer application rates and times of application.

Agronomics
In all site–years, soft red winter wheat cv. Coker 9704 was planted. Planting date, tillage system, seeding rate, subplot size, and row spacing for each experiment are detailed in Table 2. All trials followed corn and were conventionally tilled except P2001nt, where a no-till system was used. Site–years C2001 and C2002 received a preplant N application of approximately 30 kg ha–1 as N–P–K: 10–13.2–24.9% (N–P2O5–K2O: 10–30–30%), N source unknown. At P2001, P2001nt, and P2002, N25 and N30 treatments were applied as ammonium nitrate (NH4NO3: 34% N). At all other locations these treatments were applied as aqueous urea–ammonium nitrate [CO(NH2)2–NH4NO3; 30% N]. Lime and fertilizer rates other than N followed standard recommendations for North Carolina based on annual soil tests (Hardy et al., 2002; Crozier et al., 2004). Pre- and post-emergence herbicides were applied as needed (York, 2004), and weed management was excellent for all site–years except L2001, where weed populations at GS 25 were rated at approximately 22% cover.


View this table:
[in this window]
[in a new window]
  		Table 2. Site–year abbreviation (Site), planting date, tillage system used, seeding rate, subplot size, and soft red winter wheat row spacing used in this split-plot design at each location in a study of the effects of N fertilizer timing and rate on wheat grain protein variability.

 
Data Collection
The number of tillers with a minimum of three leaves in a 1-m section of row was determined at two random locations in each subplot before GS-25 N application. This resulted in 10 samples per main plot, and main plot tiller density was then estimated as the average of these samples. Subplots were harvested with a small plot Massey-Ferguson MF-8 or Gleaner K2 combine (ADCO Corp., Duluth, GA) and yields measured with a HarvestMaster grain gauge (Juniper Systems, Logan, UT). Yields were adjusted to a moisture content of 135 g kg–1. From each harvested subplot, samples of approximately 0.45 to 3.0 kg of grain were taken for grain protein and test weight analysis. Subplot grain samples of approximately 85 g were analyzed for grain N concentration using a CHN analyzer (McGeehan and Naylor, 1988) at Waters Agriculture Laboratories (Camilla, GA). Grain N concentrations were converted to grain protein by multiplying by a conversion factor of 5.83 (Kent and Evers, 1994). Test weight was determined on a volume weight basis with a DICKEY-john Grain Analysis computer (model GA C2000, DICKEY-john Corp., Auburn, IL).

Statistical Analysis
For statistical analysis, "environment" was defined as a combination of site–year and tillage system. Replications within environments and the error term for main plots were treated as random effects, while all other sources of variation and their interactions were treated as fixed effects. PROC MIXED was used to test for significant effects caused by environments, N treatments, and their interactions using SAS version 6 (SAS Institute, Cary, NC). Least-square mean separations were employed for testing differences between and among treatments. Additionally, estimates of the proportion of the total variation associated with each main effect and their interactions were determined by defining all sources of variations as random effects in PROC MIXED and then computing their estimated variance as a percentage of the total model variance.

The N25 and N30 fertilizer treatments differed in terms of "how much" and "when" N was applied. To evaluate the impact of how much N fertilizer was applied, we considered the effect of total spring N rate (rate applied at GS 25 plus that applied at GS 30) on the grain protein treatment means, and the variance computed as the SD and CV, respectively, across all seven environments. Additionally, relationships between treatment grain protein means, SD, CV, and total spring N fertilizer rates were explored using regression analysis (PROC REG, SAS version 6, SAS Institute, Cary, NC).

To elucidate the potential interaction between grain protein variability, N fertilizer treatment, and environment, we employed a method similar to that described by Eberhart and Russell (1966) for estimating a genotype's comparative yield performance across multiple environments. Over multiple environments, they regressed a genotype's yield at an environment against the mean yield of all genotypes at that environment to obtain a regression coefficient usually referred to as a "b value"; see for example Baenziger et al. (1985), Peterson et al. (1992) and Mariani et al. (1995). A genotype was considered desirable if it had a high yield and a b value close to one, meaning that it was responsive to favorable environments. An additional check of genotype desirability was a low deviation from the regression. We modified this approach using N treatments instead of genotypes, grain protein instead of yield, and with the criteria that a desirable N treatment should have low protein variability across environments. For this purpose, we assumed that b values closer to zero combined with low deviations from the regression would indicate N treatments that fostered grain protein stability across environments. PROC MIXED was used to estimate b values for each N treatment, to determine an F value to test for homogeneity of b values, to make a separation of b values by treatment, and to determine which b values were not significantly different from zero. For each N treatment, the deviations from the regression were computed as the error mean square using PROC REG.

To identify the most stable N treatments, we plotted the deviations from the regression against the b values for each N treatment. This graph was divided into quadrants based on the median values of the X and Y axes. In this graph, the stable grain protein treatments, that is, those with the lowest deviations from the regression and with b values closest to zero, fall in the lower left quadrant.

As stated above, the 25 N25 and N30 treatment combinations differed both in terms of how much N was applied (total spring N), and when that N was applied. In the quadrant chart described above, the effects of both N rate and timing could be visualized. For a more quantitative evaluation of the effect of N fertilizer timing we analyzed two subsets of the data. The first subset consisted of the five "early" N treatments that received at least 80% of the total spring N at GS 25 (i.e., N25 + N30 combinations of 33.6 + 0, 67.2 + 0, 100.8 + 0, 133.6 + 0, and 133.6 + 33.6 kg N ha–1). These data were contrasted with the second subset consisting of the five "late" treatments that received at least 80% of the total spring N at GS 30 (i.e., N25 + N30 combinations of 0 + 33.6, 0 + 67.2, 0 + 100.8, 0 + 133.6, and 33.6 + 133.6 kg N ha–1). The protein mean, SD, and regression coefficient associated with each of these 10 treatments were modeled as a function of total spring N (as a continuous variable) and fertilizer timing (i.e., early or late) as a class variable using PROC GLM.

Weather Data
Daily mean air temperature and daily total precipitation from GS 30 to harvest at each environment were obtained from the State Climate Office of North Carolina website (2004). Thirty-year daily mean temperature and precipitation data from GS 30 to harvest were obtained from the same source.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Grain Yield
Environment, N25, N30, and all their two-way interactions were significant for yield (Table 3). The N25 x N30 interaction indicated the yield response to N30 depended on the level of N25. With higher N25, there was little or no yield response to N30 compared to lower N25 (Fig. 1A ). The two-way interactions of environment with N25 and N30 indicated that yield responses to N25 or N30 depended on environment. At P2001, P2001nt, and P2002, there was little to no yield response to N25 (Fig. 1B). At C2002, yield increased and then decreased with increasing N25, while at L2001 and T2002 yield continued to increase as N25 rates increased. Similarly, the yield response to N30 differed by environment (Fig. 1C), with some environments showing little to no response (e.g., P2002), a yield plateau at higher N30 (e.g., C2002), or yields that tended to increase even through the highest N rates (e.g., T2002).


View this table:
[in this window]
[in a new window]
  		Table 3. The analysis of variance (ANOVA) for soft red winter wheat grain yield, test weight, and grain protein across seven environments (E) with N treatments applied at growth stages 25 (N25) and 30 (N30). Variability is the percentage of total model variance accounted for by each effect.

 

Figure 1
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 1. Soft red winter wheat yield response: (A) mean yield vs. N applied at growth stage (GS) 30 for five N fertilizer treatments applied at GS 25 across seven environments; (B) yield vs. N applied at GS 25 at each of seven environments; (C) yield vs. N applied at GS 30 at each of seven environments.

 
Partitioning of the total yield variance among sources showed that 68% was attributed to environment (Table 3). Only 13.4% of the variation was attributed to N treatments and their interaction (N25, N30, N25 x N30), and <7% of the variation was attributed to the interactions of environment and N treatments (E x N25, E x N30, E x N25 x N30). Clearly, environment had the strongest influence on yield variability. The highest mean yields were at P2001 and P2001nt (Table 4). These two environments also had the highest mean GS 25 tiller densities and a relatively dry spring (Table 4). In contrast, P2002 had the lowest mean yield, the lowest mean GS 25 tiller density, and an extremely wet spring. Across all environments, mean grain yield was positively correlated with mean GS 25 tiller density (r = 0.82, P < 0.05), indicating that at any given environment, mean grain yield was related to the number of tillers that had developed by GS 25. Also yield was negatively correlated to total spring precipitation (r = –0.88, P < 0.05), indicating that wetter environments had lower yields.


View this table:
[in this window]
[in a new window]
  		Table 4. Overall environmental means for yield, test weight, grain protein, tiller density, mean daily temperature, and daily total precipitation from growth stage (GS) 30 to harvest in a study of the effects of N fertilizer timing and rate on wheat grain protein variability. The mean values of daily temperature and daily total precipitation for the preceding 30 yr at each environment (E) are also shown.

 
Test Weight
Environment, N25, N30, and their interactions were all significant for test weight with the exception of the environment x N25 interaction (Table 3 and Fig. 2 ). The three-way interaction indicated that test weight response to any treatment factor depended on the levels of the other treatment factors (Fig. 2). However, partitioning the sources of variation showed that 90.5% of the variability in test weight was attributed to environment, and only about 2.0% of the variation was attributed to N treatments and their interactions with environment (Table 3). A test weight of 747 kg m–3 or above is considered representative of good soft wheat grain quality (USDA–ARS Soft Wheat Quality Lab., 2004), and producers can be penalized financially when test weights are below this standard. Mean test weights at C2002, P2002, and T2002 were below this standard; these sites had cool and extremely wet springs (total precipitation above the 30-yr mean, Table 4). Mean test weight was not correlated with mean yield or grain protein. The highest mean test weights occurred at C2001 and L2001 where total spring precipitation levels were slightly below the 30-yr mean (Table 4).


Figure 2
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 2. Soft red winter wheat test weight response to N applied at growth stage (GS) 30 for five N fertilizer treatments applied at GS 25 at each of seven environments: (A) C2001, (B) C2002, (C) L2001, (D) P2001, (E) P2001nt, (F) P2002, (G) T2002.

 
Grain Protein
Environment, N25, and N30 and all their interactions had a significant effect on grain protein (Table 3), indicating that grain protein response to a treatment factor depended on the levels of both the other treatment factors as seen in Fig. 3 . In several environments (e.g., C2001 and C2002) grain protein at the lower N25 rates, initially decreased as N30 increased. This is a predictable response in grain protein accumulation in low N environments (Gooding and Davis, 1997). This response was not seen at P2001 and P2001nt, indicating a possibility of higher soil residual N. At T2002, there was a plateau in grain protein response to high N25 and N30 combinations, while at other locations grain protein continued to increase at higher N rates (Fig. 3).


Figure 3
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 3. Soft red winter wheat grain protein response to N applied at growth stage (GS) 30 for five N fertilizer treatments applied at GS 25 at each of seven environments: (A) C2001, (B) C2002, (C) L2001, (D) P2001, (E) P2001nt, (F) P2002, (G) T2002.

 
Only 23.3% of the variability in grain protein was attributed to environment (Table 3). There were no direct relationships between grain protein and daily average temperature or daily total precipitation from GS 30 to grain fill (Table 4). Grain protein was not related to yield, test weight, or tiller density. The majority (51.4%) of the variability in grain protein was attributed to N treatments (N25, N30, N25 x N30; Table 3), with an additional 7.6% attributed to the interactions of N treatments and environment (E x N25, E x N30, E x N25 x N30). The contrast between grain protein and test weight regarding the proportion of response attributable to N and to environment is apparent in Fig. 3 (grain protein) vs. Fig. 2 (test weight).

Total Spring Nitrogen Applied
Across all N25 and N30 treatment combinations, the total amount of spring N applied ranged from 0.0 to 268.8 kg ha–1, and the 25 treatment grain protein means (averaged across environments) ranged from 104.0 to 138.5 g kg–1 (Table 5). There was a strong positive relationship (r2 = 0.97, Fig. 4A ) between treatment mean grain protein and the total amount of spring N applied by that treatment. Treatment mean grain protein was unaffected by the first N increment (33.6 kg N ha–1), but then showed a near-linear positive response for higher N rates.


View this table:
[in this window]
[in a new window]
  		Table 5. Total spring N applied, N applied at growth stage 25 (N25) and 30 (N30), and the resulting treatment grain protein mean (averaged across environments), SD and CV. The deviations from the regression of N treatment grain protein content at a given environment against the mean grain protein content of all treatments at that environment, and the resultant regression coefficient (b value) for each fertilizer N treatment are also shown for this study of the effects of N fertilizer timing and rate on wheat grain protein variability.

 

Figure 4
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 4. (A) Treatment mean wheat grain protein content (averaged across environments); and (B) treatment grain protein SD (solid black circles), and CV (open gray circles) vs. the total amount of spring N applied for each of 25 N fertilizer treatment combinations.

 
Higher spring N rates not only resulted in higher treatment mean grain protein (Fig. 4A), but also resulted in higher grain protein variability as measured by the fertilizer treatment SD and CV (Fig. 4B and Table 5). Treatment mean grain protein content was positively correlated with both SD and CV (r = 0.81 and 0.57, respectively). The relationships in Fig. 4A and 4B demonstrated that as total spring N increased, treatment mean grain protein also increased but at the cost of higher variability across environments.

In the ANOVA for grain protein, all interaction terms involving environment and N treatment were significant (Table 3). Figure 5 illustrates this interaction of N treatment and environment for a representative set of N treatments. For each N treatment, grain protein at a given environment (Fig. 5, y axis) was regressed against the protein mean of all treatments at that environment (Fig. 5, x axis). The regression coefficient (b value) and deviations from the regression were then calculated (Table 5). An F-test for differences among these b values was significant (P = 0.0001), and all b values were significantly different from zero. Low total spring N fertilizer rates tended to have b values closer to zero; as N inputs increased, the b values also increased (Table 5, Fig. 5). For example, the two N treatments that applied only 33.6 kg N ha–1 had b values of 0.37 and 0.66. At the other extreme, the highest N fertilizer rate (268.8 kg N ha–1) had a b value of 1.60. Overall, the treatment b values were linearly related to the total amount of N applied (b value = 0.0051Ntot + 0.37, r2 = 0.88, P < 0.05, where Ntot is the total amount of N applied).


Figure 5
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 5. Linear regressions of grain protein content (g kg–1) for individual treatments at a given environment against the mean grain protein content (g kg–1) of all treatments at that environment. Each of the seven environments is identified along the x axis (T2002; Tidewater Research Station 2002, C2001; Cunningham Research Station 2001, L2001; Lower Coastal Plains Research Station 2001, C2002; Cunningham Research Station 2002, P2002; Piedmont Research Station 2002, P2001; Piedmont Research Station 2001, and P2001nt; Piedmont Research Station 2001 no-till). Four N fertilizer treatments (those receiving a total of 0.0, 67.2, 134.4, and 201.6 kg spring N ha–1 with half the N applied at growth stage 25 and the second half at growth stage 30) are shown.

 
As noted above, as the total amount of spring N increased, the treatment mean grain protein content (across all environments) and SD also increased (Fig. 4A and B). The fertilizer treatment b values and SD were also linearly related (b value = 0.15SD- 0.21, r2 = 0.82, P < 0.05). Apparently, grain protein was unresponsive to environments at low N rates, and therefore relatively stable. At high N rates, however, there was a large difference between the protein content produced at environments with high protein potential compared to that produced at low protein potential environments, which caused the high SD and CV associated with higher N rates. These data indicate that when high spring N rates are used in disparate environments, high protein grain may be produced at some locations, but those N rates will also result in substantial protein variability across environments.

To stabilize regional grain protein content, an ideal N fertilizer treatment would have a low b value and a low deviation from the regression. To identify which N treatments best met these criteria, treatment deviations from the regression were plotted against their associated b values (Fig. 6 ) and this graph was then divided into quadrants. Treatments with the lowest deviations from the regression and the lowest b values are by definition in the lower left quadrant of such a figure. Using this approach, seven N25 + N30 treatment combinations were identified as being potentially ideal for protein stability; 33.6 + 0.0, 67.2 + 0.0, 134.4 + 0.0, 67.2 + 33.6, 67.2 + 67.2, 33.6 + 100.8, and 0.0 + 100.8 kg N ha–1.


Figure 6
View larger version (12K):
[in this window]
[in a new window]
  		Fig. 6. Quadrant chart with deviations from the regression (of N treatment grain protein content at a given environment against the mean grain protein content of all treatments at that environment) on the y axis (computed as the error mean square), and the regression coefficient (b value) on the x axis for each of the 25 N treatment combinations. The amount of N fertilizer applied at growth stage 25 and 30 (rounded to the nearest kg ha–1) are indicated for the treatment combinations falling in the lower left quadrant.

 
Nitrogen Timing
Five of the seven N treatments that appeared to be ideal for reducing regional grain protein variability (Fig. 6) had at least 50% of the N applied at GS 25. This suggested that the timing of the N application was as important as the fertilizer N rate. To test this timing effect, we analyzed two subsets of the data. The first subset consisted of the five "early" treatments that received at least 80% of the total spring N at GS 25. These data were contrasted with the second subset consisting of the five "late" treatments that received at least 80% of the total spring N at GS 30. For both the treatment SD and the b values, fertilizer timing ("early" or "late") was statistically significant as a class variable, and the linear term for total spring N applied was a statistically significant co-variable (Table 6). On average, at any given total spring N rate, applying ≥80% of the total N at GS 25 resulted in a SD that was 2.1 g kg–1 lower, and a b value that was 0.21 lower compared to applying ≥80% of the total N at GS 30. Applying the majority of N at GS 25 reduced overall variability and resulted in more stable N treatments.


View this table:
[in this window]
[in a new window]
  		Table 6. Modeling treatment grain protein SD, b values (b), and mean (P) as a linear function of total amount of spring N fertilizer applied (Ntot) and when that fertilizer was applied (Timing). Timing was defined as "early" if at least 80% of Ntot was applied at growth stage 25, and "late" if at least 80% of Ntot was applied at growth stage 30.

 
Timing of N application also had an impact on the resultant grain protein. For a given amount of total spring N fertilizer, applying ≥80% of that fertilizer at GS 25 resulted in a statistically significant elevation in grain protein content compared to applying it at GS 30 (Table 6). While this increase was statistically significant, it was small (3.3 g kg–1) and consequently probably of little agronomic importance.


   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Our results indicated that most of the variability in yield was due to environment. Many environmental factors including timeliness of planting (Knapp and Knapp, 1978), reduced tillering (Weisz et al., 2001), and weed competition (Gooding and Davis, 1997) can contribute to yield variability. Wheat was planted late at P2002 and T2002, and late planting coupled with cool temperatures resulted in low tiller densities that may explain the lower yields at those environments. High weed populations at L2001 may have negatively affected tiller density and yield at that location. Overall, GS-25 tiller density and spring precipitation appeared to be the primary environmental factors influencing yield variability.

The majority of the variability in test weight was also attributed to environment. Environments P2002 and T2002 had the lowest mean test weights and the highest spring precipitation. This was consistent with reduced test weights being associated with environments that had an increased chance of grain wetting during the formation or filling process. Schuler et al. (1994), Weisz and Bowman (1999), and Yamazaki and Briggle (1969) reported that environment was a major contributor to variability in test weight of soft red winter wheat grown in temperate climates.

In contrast to grain yield and test weight, the majority of protein variability (51.4%) was attributed to N treatments. The increase in grain protein with spring-applied N is consistent with findings by researchers in other wheat production regions (Johnson et al., 1973; Fowler et al., 1989; Smith and Gooding, 1996; López-Bellido et al., 1998). Clearly, if producers within a region use different N rates, that fact alone will result in variability in soft red winter wheat grain protein content.

Grain protein content variability across environments also increased with higher N rates (Fig. 4B). In our study, all the interaction terms that included environment and N treatment were significant for grain protein. Low N rates resulted in low grain protein levels that were relatively stable across environments (Fig. 5). At environments with low mean protein, higher N rates had a relatively small effect on protein. However, high N rates in responsive environments resulted in large increases in protein (Fig. 5). Consequently, if high N rates are applied throughout a region this could result in a wide range of grain protein levels.

Based on work in other wheat production areas (Vaughan et al., 1990; Stark and Tindall, 1992; Kelley, 1995; Woodard and Bly, 1998) we hypothesized that grain protein content might be higher when N was delayed until GS 30. However, we observed a small decrease in grain protein content when application was delayed to GS 30. Furthermore, there was a difference in protein stability between treatments that applied ≥80% of spring N at GS 25 instead of GS 30. There was a trend for the b values associated with the five early N treatments to be lower than those associated with the five late treatments (Table 5). When the early and late treatments were pooled into two groups, this trend was statistically significant (Table 6), which is perhaps the most interesting and important of our findings. At a given N rate, applying the majority of that N at GS 25 resulted in a slightly higher grain protein content, but a lower b value, and consequently a protein content that was less sensitive to environmental differences and more regionally stable.

Our secondary objective was to determine if there were N fertilizer recommendations that might minimize regional grain protein variation for soft red winter wheat intended for the baking industry. Some of the 25 treatments explored in this study did result in lower protein variability. Based on the criteria of low deviations from the regression and a low b value, seven N25 + N30 treatment combinations (33.6 + 0.0, 67.2 + 0.0, 134.4 + 0.0, 67.2 + 33.6, 67.2 + 67.2, 33.6 + 100.8, and 0.0 + 100.8 kg N ha–1) were identified as the most stable (Fig. 6). These treatment combinations also had relatively low SDs and CV (Table 5). Of these N treatments, those that apply 67.2 kg N ha–1 or less would most likely not be agronomically feasible, but the remaining five treatments comprise N rates that would generally optimize yield while minimizing regional protein variability.

Our results suggest some general recommendations that might lead to lower regional grain protein variability. The first recommendation is to reduce the range (45–202 kg ha–1) of N fertilizer rates used across the region. One of the biggest contributors to high protein variability in this study was high N fertilizer rates. Consequently, the second recommendation is to avoid overapplication of N beyond what is required to optimize yield and economic return. Current recommendations in North Carolina are that spring N rates not exceed 134 kg ha–1 (Weisz and Heiniger, 2004). Applying 100.8 to 134.4 kg N ha–1 resulted in greater grain protein stability than 168 to 269 kg N ha–1. Limiting N application rates to 101 to 134 kg N ha–1 would reduce the regional protein variability compared to the range in rates currently used. Scharf and Alley (1993) proposed using a GS 30 tissue test to optimize spring N fertilizer rates in the southeastern USA. This technique is a good method for optimizing wheat yields and minimizing the use of excessively high or low N fertilizer rates. The third recommendation to reduce protein variability is to apply spring N at GS 25 and avoid waiting until later in the season. Five of the seven treatment combinations identified as most stable for grain protein had at least 50% of the total spring N applied at GS 25, and "early" N applications increased stability compared to "late" ones. In essence, regional interest would be served well by reducing the range of N rates applied, realistically applying N based on yield potentials or an in-season tissue test, and avoiding later N applications.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
This research was sponsored in part by Initiative for Future Agriculture and Food Systems Grant no. 00-52103-9644 from the USDA Cooperative State Research, Education, and Extension Service.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Agricola
Right arrow Articles by Farrer, D. C.
Right arrow Articles by White, J. G.
Related Collections
Right arrow Other Crop Management
Right arrow Best Management Practices
Right arrow Wheat

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome