Grain protein as a post-harvest index of N sufficiency for hard red spring wheat in the semiarid prairies F. Selles, and R. P. Zentner Semiarid Prairie Agricultural Research Centre, Agriculture and Agri-Food Canada, Swift Current, Saskatchewan, Canada S9H 3X2. Received 27 July 2000, accepted 10 May 2001. Selles, F. and Zentner, R. P. 2001. Grain protein as a post-harvest index of N sufficiency for hard red spring wheat in the semiarid prairies. Can. J. Plant Sci. 81: 631 636. Results from fertilizer trials with hard red spring wheat (Triticum aestivum L.) conducted throughout southwestern Saskatchewan under fallow and cereal stubble cropping conditions were used to determine if grain protein concentration (GPC) could be used as an index of N sufficiency to the crop. Critical GPC were determined using the Cate-Nelson R 2 procedure. Grain yield and protein concentration were negatively correlated under stubble and for fallow cropping when yields were below 2858 kg ha 1 ± 179, with grain protein decreasing by 15 mg g 1 for every 1000 kg ha 1 yield increase. In these two groups of observations, water and N availability, N yield and grain produced per unit N available suggested that water availability was the dominant factor limiting grain yield. For the portion of fallow observations in which grain yields were higher than 2858 kg ha 1, water availability was not limiting, and N availability controlled grain yield and protein concentration. In this group, a GPC of 128 mg g 1 (range of 123 to 135 mg g 1 ) marked the transition between N deficiency and sufficiency. Under stubble cropping and for the lower-yielding portion of the fallow cropping system, where water stress was predominant, the Cate- Nelson analysis identified critical protein concentrations of 160 and 154 mg g 1, respectively. However, these critical concentrations separated populations into moderately and severely water-stressed crops, rather than providing a separation based on N availability. We concluded that GPC as a post-harvest index of N sufficiency must be used with caution in southwestern Saskatchewan. Grain protein concentration below the critical limit of 128 mg g 1 is a reliable indicator of low N sufficiency, but high grain protein does not necessarily imply N sufficiency because, frequently, grain yield and protein concentration are negatively correlated due to water stress. Key words: Yield, protein, N availability, critical levels, water stress Selles, F. et Zentner, R. P. 2001. Utilisation de la concentration de protéines dans le grain comme index post-messianique de l apport de N pour le blé roux vitreux de printemps cultivé dans les prairies semi-arides. Can. J. Plant Sci. 81: 631 636. Les auteurs ont recouru aux résultats des essais de fertilisation sur le blé roux vitreux de printemps (Triticum aestivum L.) cultivé sur jachère et sur chaume de céréale dans le sud-ouest de la Saskatchewan pour voir si la concentration de protéines dans le grain peut servir à déterminer si la culture reçoit assez de N. La concentration critique de protéines dans le grain a été établie par la méthode R 2 de Cate-Nelson. Le rendement grainier et la concentration de protéines présentent un corrélation négative pour la culture sur chaume et pour un rendement inférieur à 2 858 kg ha 1 ± 179 dans le cas de la culture sur jachère, la concentration de protéines diminuant de 15 mg g 1 pour chaque hausse de rendement de 1000 kg ha 1. Dans les deux séries d observations, la disponibilité d eau et d azote, le rendement azoté et le grain produit par unité N laissent croire que le principal facteur limitant du rendement grainier est la quantité d eau disponible. Quand le rendement de la culture sur jachère dépasse 2 858 kg ha 1, la quantité d eau disponible n est plus un facteur limitant, mais la disponibilité de l azote régule le rendement grainier et la concentration de protéines. Dans cette série d observations, la concentration de 128 mg g 1 de protéines dans le grain (intervalle de 123 à 135 mg g 1 ) marque la transition entre un apport suffisant de N et la carence. L analyse de Cate-Nelson fixe la concentration critique de protéines respectivement à 160 et à 154 mg g 1 pour la culture sur chaume et pour la culture sur jachère à rendement moins élevé, où prédomine le stress hydrique. Néanmoins, les seuils critiques permettent de séparer les peuplements en fonction du stress hydrique (modéré ou grave) plutôt que d après la disponibilité d azote. On en conclut qu il faut faire preuve de prudence quand on utilise la concentration de protéines dans le grain comme index pour établir si l apport de N est suffisant, dans le sud-ouest de la Saskatchewan. Une concentration de protéines inférieure au seuil de 128 mg g -1 est un indice fiable d une quantité limite de N, mais une forte concentration ne traduit pas nécessairement un apport suffisant de N, car le rendement grainier et la concentration de protéines présentent souvent une corrélation négative à cause du stress hydrique. Mots clés: Rendement, protéine, disponibilité d azote, seuil critique, stress hydrique Grain protein is an essential factor determining the breadmaking quality of wheat (Triticum aestivum L.) (Lukow and Preston 1998). To ensure adequate supplies of high protein wheat, the Canadian Wheat Board introduced a schedule of protein price premiums for bread and durum wheats (Smith et al. 1998). In response to these higher potential economic returns, grain producers have become increasingly interested in optimizing N fertilization of their wheat crops. In environments were N limits plant growth, N fertilization is 631 essential for improving grain yields and protein concentrations. The response of crops to fertilization follows the law of diminishing returns, with yield increases becoming progressively smaller with each extra N addition, until finally yield increases cease and the crop reaches its maximum yield (Selles et al. 1992). Additions of extra N above those Abbreviations: GPC, grain protein concentration; YNUE, amount of grain produced per unit available N
632 CANADIAN JOURNAL OF PLANT SCIENCE required for maximum yield may lead to yield decreases (Kramer 1979), but invariably they result in increased GPC (Benzian and Lane 1981) because at high levels of N availability photosynthesis remains nearly constant while N uptake continues to increase (Lawlor et al. 1989). Some researchers have proposed that GPC, as a complement to soil testing for N, could be a useful indicator of N sufficiency for optimum yield of wheat (Grant and Flaten 1998). Goos et al. (1982), using a chi-square test, determined that winter wheat yields were often depressed by low N availability when GPC was 115 mg g 1 or less. For the eastern Canadian prairies, it was estimated that a GPC of 135 mg g 1 appears to indicate the critical level of N sufficiency required for optimum yields of hard red spring wheat (Flaten and Racz 1997). In Montana, Long and Engel (1998) also suggested that 135 mg g 1 was the critical level to separate wheat into N-deficient or N-sufficient classes. For wheat grown in northern Europe, 122 mg g 1 was identified as the critical level (Virtanen and Peltonen 1996). The objective of this study was to determine if GPC of wheat can be used as an index of N sufficiency for wheat grown on fallow and stubble in the semiarid region of southwestern Saskatchewan. For this, we used average yield and protein responses to N fertilization obtained from fertilizer trials conducted on various soil types throughout southwestern Saskatchewan during the 1967 to 1999 period. MATERIALS AND METHODS Fertilizer response tests for hard red spring wheat were conducted by the Semiarid Prairie Agricultural Research Centre (SPARC) at Swift Current on up to 15 sites annually throughout southwestern Saskatchewan, in soils with textures ranging from sandy loam to heavy clay. Under fallow cropping conditions, a total of 256 replicated tests were conducted from 1967 to 1993 and from 1996 to 1999, distributed across 126 soil-year combinations, or environments, with a total of 2689 means of environment-fertilizer rate combinations. Under cereal stubble cropping conditions, 156 replicated tests were conducted from 1967 to 1971, from 1978 to 1993, and from 1996 to 1999, distributed across 96 soil-year combinations with a total of 1828 means of soil-year-fertilizer rate combinations. Only the averages of environment-fertilizer combinations were used in this study. Details of the methods and procedures used in the field tests are described in Selles et al. (1992). For potential evaporation, data from Class A pan evaporation measured at the SPARC meteorological station was used. This is the closest station to all field sites that records pan evaporation. Separation of samples into protein limited and non-limited classes was carried out using two procedures; the chisquare test suggested by Goos et al. (1982), and the R 2 method proposed by Cate and Nelson (1971). The chi-square method consists of classifying yields into high and low yield classes. The high yield class consisted of yields that were not significantly (P > 0.05) different from the maximum yield for each individual test. The low class yield consisted of all yields that were significantly lower than those of the maximum yield for each test. The data were further divided into low and high protein classes based on arbitrary protein cutoff points set between 90 and 190 mg g 1 at 2.5 mg g 1 protein intervals. At each cutoff point, the data can be summarized in a 2 2 contingency table and one can calculate a chi-square statistic for the contingency table. In this study, we used a chisquare test with correction for continuity, as defined by Steel and Torrie (1960) [formula 19.10, page 371], given that this test has a single degree of freedom. When the protein cutoff is set either low or high, one of the diagonal elements of the contingency table will be larger than the other, leading to a large chi-square statistic. Large chi-square values indicate a significant interaction between the two classifications, and that the protein classification can be used to predict the yield classification, or vice-versa. As the protein cutoff value is moved towards mid-range values, the chi-square statistic decreases, indicating that samples within this region are homogeneous, and cannot be separated into discrete classes. Keisling and Mullinix (1979) stated that this indicates failure of the indicator variable (GPC in our study) to discriminate for classification purposes. Thus, this region constitutes a transition zone, or range, beyond which the samples change from protein deficient to sufficient, or vice-versa. The Cate and Nelson (1971) procedure consists of organizing the data into an ordered array of X (GPC) and Y values (grain yield), with the X, Y pairs maintained in this order throughout the analyses. Starting with a protein cutoff value that places at least two X, Y pairs in the low protein class, the samples are divided into low and high protein classes. At each point, an R 2 value is calculated as: 1 (CSS l + CSS h )/CSS t where CSS l is the corrected sum of squares of grain yield for the low protein class; CSS h is the corrected sum of squares of grain yield for the high protein class; and CSS t is the corrected sum of squares of grain yield for the whole data set. The critical, or transition GPC value is that at which the R 2 reaches a maximum. This procedure has been used to separate cereal yield classes, based on their GPC (Virtanen and Peltonen 1996). Statistical Analyses All statistical analyses were conducted with JMP v 3.2.6 (SAS Institute, Inc. 1994). Chi-square and Cate-Nelson R 2 values for each protein cutoff point were calculated with a custom designed computer program. The probability of no interaction for each chi-square value was calculated with the chi-square distribution function of the calculator in JMP 3.2.6 (SAS Institute, Inc. 1994). Mean separations for variables in the classes separated in this study were estimated with the Tukey-Kramer HSD (honestly significant difference) test (Kramer 1956). RESULTS AND DISCUSSION Yield-protein Relationships There was a negative yield-protein relationship under stubble cropping conditions (Fig. 1). A linear regression of GPC
SELLES AND ZENTNER POST-HARVEST EVALUATION OF N AVAILABILITY 633 Fig. 1. Yield-protein relationships for wheat grown on cereal stubble (a) and fallow (b). on grain yield revealed that GPC decreased by 15 mg g 1 (P 0.05) for every 1000 kg ha 1 increase in grain yield (R 2 = 0.21, P < 0.0001). Under fallow cropping, however, the yield-protein relationship was negative at low yield levels, but it became positive as grain yield increased beyond a threshold level (Fig 1). A spline regression for a function with two linear segments linked at a common point (knot) was fitted to these data using a non-linear regression procedure (Freund and Littell 1986). This function (R 2 = 0.28, P < 0.001) revealed that the threshold yield (or knot) at which the slope of the response changed from negative to positive, was 2858 kg ha 1 (asymptotic 95% confidence interval 2652 to 3010 kg ha 1 ). Below this threshold yield, GPC behaved as it did for the stubble system, decreasing by 15 mg g 1 for every 1000 kg ha 1 yield increase. At yields above the threshold, GPC increased by 11 mg g 1 with every 1000 kg ha 1 increase in grain yield. Negative yield-protein relationships normally happen when growth factors other than N availability (e.g., water) limit grain yield, especially when N supply is relatively constant (Terman 1979; Clarke et al. 1990). Conversely, in environments where N availability controls or limits plant growth, grain yield and GPC increase when N supply improves, leading to a positive relationship (Gauer et al. 1992). We tested this hypothesis by inspecting the average water and N supply to the crop for the negative and positive domains of the yield-protein relationship, using the yield threshold estimated by the spline regression as the cutoff point. Although the total amount of water received by both domains was the same, the negative domain of the fallow system had significantly more available water in the soil (P 0.05), but received less rainfall (P 0.05) and was subjected to higher evaporation (P 0.05) than the positive domain (Table 1). These differences in the temporal distribution of water stress are extremely important in determining or modifying yield potential or GPC. Under the conditions experienced by the negative domain, and especially with abundant N supply, cereal crops tend to lower grain yields and increase GPC as result of haying off (Campbell et al. 1977; van Herwaarden et al. 1998). Expressing the total amount of available water as a fraction of potential evaporation, or as a water deficit factor, calculated as available water minus potential evaporation, provides further evidence of the substantial water stress affecting the crop in the negative domain (Table 1). Although the amount of available N (soil NO 3 -N + fertilizer N) was lower (P 0.05) in the negative domain, crop growth in this domain was limited by water availability rather than by N availability, as evidenced by the significantly lower (P 0.05) amount of grain produced per unit available N (YNUE) (Table 1). Previous studies demonstrated that the magnitude of YNUE was directly proportional to water availability (Clarke et al. 1990). This provides further evidence, and allowed us to conclude that the crop in the negative yield-protein relationship domain was under substantially higher water stress than the crop in the positive domain. Under stubble cropping, the water stress was as intense as in the negative yield-protein relationship domain of the fallow system when assessed by the water deficit term (Table 2). Improvements in water
634 CANADIAN JOURNAL OF PLANT SCIENCE Table 1. Least squares means for water and N supply to wheat grown on fallow, separated according to the nature of the yield-protein relationship Positive Negative Variable Units YPR z YPR Spring soil water 0 120 cm mm 102a 126b 1 May 31 July rainfall mm 165b 141a 1 May 31 July evaporation y mm 570a 718b Water deficitx mm 302a 448b AW/EvapW 0.48b 0.39a AVN v kg ha 1 87b 75a NUE u kg kg 1 45b 28a z Data separated into groups according to nature of the yield-protein relationship (YPR). y Class A pan evaporation at Swift Current. x (Spring soil water + 1 May 31 July rainfall) potential evaporation. w Available water as fraction of potential evaporation. v Soil NO 3 -N (0 60 cm) + fertilizer N. u Grain produced per unit available N. a, b Values within a row followed by the same letter are not significantly different (P > 0.05) based on the Tukey-Kramer HSD test. availability lead to higher yields, which diluted GPC leading to a negative yield-protein relationship throughout the range of grain yields observed. Separation of Domains by Grain Protein Determination of the critical concentrations of protein for wheat grown on fallow was done separately for the two yield domains as defined by the nature of their yieldprotein relationships. For wheat grown on stubble, the critical protein concentration was determined for the entire data set as the yield-protein relationship was negative for all observations. For wheat grown on stubble, the chi-square test showed a maximum probability of no interaction at a critical protein concentration of 159 mg g 1 with a transition range extending from 148 mg g 1 to 166 mg g 1 (data not shown). This transition range, the region where the classifying variable is unable to separate classes, was extremely wide encompassing nearly one-quarter (22.4%) of all observations in the stubble system. For wheat grown on fallow, the chi-square test failed to separate the grain yield observations according to their GPC content, for either domain of the yield-protein relationship. These latter results would suggest that GPC and grain yields are independent, in contradiction to the significant spline function fitted earlier to these data (R 2 = 0.28, P < 0.0001). The Cate-Nelson R 2 test was more sensitive, and was able to identify clear critical protein concentrations for all the data domains used in this study. For the positive domain of the fallow system, the R 2 procedure identified a broad peak extending from 123 to 135 mg g 1 with three nearly identical peaks at 124, 127, and 132 mg g 1 (R 2 = 0.15 each peak) at the summit. We estimated the cutoff point as the midpoint of the three peaks (128 mg g 1 ) and the transition range, that is the GPC region where samples cannot be classified into N-insufficient or N-sufficient classes, as the breadth of the main peak (122 to 135 mg g 1 ) (Fig. 2). This critical value was within the range of GPC identified in other studies as indicators of N-insufficiency for full expression of yield potential (Goos et al. 1982; Virtanen and Peltonen 1996; Flaten and Racz 1997; Long and Engel 1998). Further, this value is remarkably close to the average GPC of the statutory check used in the breeding of hard red spring wheat cultivars (T.N. McCaig, personal communication, SPARC, Swift Current, SK). Separating this domain into low and high GPC groups, using the 128 mg g 1 cutoff revealed that water availability and potential evaporation were similar (P > 0.05) for both groups (Table 2). Available N and N yield, however, were significantly higher (P 0.05) in the high GPC group than in the low GPC group, indicating that Table 2. Least squares means of grain yield, protein concentration, and water and N supply for fallow and stubble grown wheat, separated into domains by grain protein as classing variable Fallow positive YPR z Fallow negative YPR Stubble High Low High Low High Low Variable Units protein protein protein protein protein protein Critical protein concentration mg g 1 128 154 160 Grain yield kg ha 1 3505e 3274d 1499b 2008c 907a 1467b GPC mg g 1 142 119 166 134 172 130 N yield kg ha 1 87f 68e 43c 47d 27a 33b Spring soil water (0 120 cm) mm 104b 100b 125c 127c 62a 76b 1 May 31 July rainfall mm 168c 160bc 130a 146b 129a 151b 1 May 31 July evaporation y mm 575a 561a 763d 699c 764d 665b Water deficit x mm 303a 301a 506c 423b 572d 423b AVW/Evap w 0.48e 0.47e 0.35b 0.41d 0.26a 0.38c AVN v kg ha 1 122d 92c 86c 73b 86c 64a NUE u kg kg 1 34e 40f 19b 31d 12a 28c n 227 112 708 1642 396 1432 z Fallow data separated into groups according to nature of the yield-protein relationship (YPR). y Class A pan evaporation at Swift Current. x (Spring soil water + 1 May to 31 July rainfall) minus potential evaporation. w Available water as fraction of potential evaporation. v Soil NO 3 -N (0-60 cm) + fertilizer N. u Grain produced per unit available N. a f Values within a row followed by the same letter are not significantly different (P > 0.05) based on the Tukey-Kramer HSD test.
SELLES AND ZENTNER POST-HARVEST EVALUATION OF N AVAILABILITY 635 in the positive domain of wheat grown on fallow, N availability was the main factor determining grain yield and GPC. These results confirm that in the semiarid prairies when water availability does not restrict crop growth, a GPC of 128 mg g 1 (range of 123 to 135 mg g 1 ) marks the transition between N-deficiency and N-sufficiency, in agreement with the results of studies conducted in moister environments (Goos et al. 1982; Virtanen and Peltonen 1996; Flaten and Racz 1997; Long and Engel 1998). In the negative domain of wheat grown on fallow, the R 2 test identified a narrow peak extending from a GPC value of 152 to 158 mg g 1 and maximum R 2 at 154 mg g 1 (R 2 = 0.50) (Fig. 2). For the stubble system, this test identified a peak with broad summit extending from 143 to 167 mg g 1 and a maximum R 2 at 160 mg g 1 (R 2 = 0.28). This critical value is very close to that determined for the negative domain of the fallow system, and remarkably close to the critical value identified earlier by the chi-square test (159 mg g 1 ) (Fig 2). The critical protein concentrations for the observations with a negative yield-protein relationship (fallow system 154 mg g 1, stubble system 160 mg g 1 ) were high, and well within a range of GPC identified in previous studies as N- sufficient (Flaten and Racz 1997; Long and Engel 1998). More importantly, the critical values were within the range of N sufficiency determined in this study for wheat grown under fallow with low water stress. Furthermore, when the observations with negative yield-protein relationship were separated into high and low GPC ranges, according to their critical values, they revealed that the mean GPC of each of the ranges was substantially higher than the GPC critical value determined for the positive domain of the fallow system (Table 2). Grain and N yields in the low protein ranges were significantly larger (P 0.05) than in their respective high protein ranges (Table 2). These observations suggest that N availability for these crops was not limiting growth. Inspection of indicators of water stress such as potential evaporation, the ratio of available water to evaporation, and the water deficit factor, however, suggested that observations Fig. 2. Results of the Cate- Nelson R 2 analysis, showing critical protein values (vertical lines) for positive yield-protein relationship domain (YPR) of wheat grown on fallow (a), negative yield-protein relationship domain for wheat grown on fallow (b), and wheat grown on stubble (c). in the high GPC ranges were subjected to more severe water stress than those in their respective low GPC ranges. Furthermore, grain produced per unit N in the water stressed high GPC range was significantly lower than in the low range, supporting the evidence for water stress induced high GPC (Clarke et al. 1990). We hypothesized then, that the GPC critical value for these data marks the transition between a high protein range, characterized by severe limitations to crop growth due to severe water stress, and a low protein range, characterized by moderate growth limitations due to water stress. As such, in crops subjected to considerable water stress, GPC is not a useful indicator of N sufficiency. CONCLUSIONS Results from this study indicate that under the limited available water conditions of the semiarid prairies, water stress regularly restricts grain yields of wheat grown on stubble and on fallow, leading to a negative yield-protein relationship. Under these conditions, GPC in itself does not provide a reliable index of N-sufficiency for the crop. For wheat grown on fallow in years when water stress is less severe, however, the GPC provides a useful guide to judge the N sufficiency of the crop. Under these conditions, a GPC of 128 mg g 1 marks the separation between crops deficient and sufficient in N for realization of their full yield potential. However, this was true for only 12.6% of the fallow observations, and only 7.5% of all observations in this study. Grain protein concentration values below the critical value invariably indicate that the crops were grown with inadequate N supply to achieve their yield potential. However, GPC values above the critical level may not necessarily be interpreted as adequate N supply levels to the crops, because high GPC may arise from adequate N supplies, or from water stress that limits grain production. Use of GPC as an index of N availability in the semiarid environment of southwestern Saskatchewan must be interpreted carefully, and supported by additional information because water stress often is the most important factor determining grain yield and GPC.
636 CANADIAN JOURNAL OF PLANT SCIENCE ACKNOWLEDGMENT We gratefully acknowledge the technical assistance of D. C. James, S. Fleck, and W. Chalk Benzian, B. and Lane, P. 1981. Interrelationships between nitrogen concentration in grain, grain yield and added fertiliser nitrogen in wheat experiments in south-east England. J. Sci. Food. Agric. 32: 35 43. Campbell, C. A. Cameron, D. R., Nicholaichuk, W. and Davidson, H. R. 1977. Effect of fertilizer N and soil moisture on growth, N content, and moisture use by spring wheat. Can. J. Soil Sci. 57: 289 310. Cate, R. B., Jr. and Nelson, L. A. 1971. A simple statistical procedure for partitioning soil test correlation data into two classes. Soil Sci. Soc. Am. Proc. 35: 658 660. Clarke, J. M., Campbell, C. A., Cutforth, H. W., DePauw, R. M. and Winkleman, G. E. 1990. Nitrogen and phosphorus uptake, translocation, and utilization efficiency of wheat in relation to environment and cultivar yield and protein levels. Can. J. Plant Sci. 70: 965 977. Flaten, D. N. and Racz, G. J. 1997. Nitrogen fertility and protein in red spring wheat. Pages 72 75 in Proc. Manitoba Agri-Forum. 18 February 1997. Winnipeg, MB. Freund, R. J. and Littell, R. C. 1986. SAS system for regression, 1986 ed. SAS Institute Inc., Cary. NC. Gauer, L. E., Grant, C. A., Gehl, D. T. and Bailey, L. D. 1992. Effects of nitrogen fertlization on grain protein content, nitrogen uptake, and nitrogen use efficiency of six spring wheat (Triticum aestivum L.) cultivars, in relation to estimated moisture supply. Can. J. Plant Sci. 72: 235 241. Goos, R. J., Westfall, D. G., Ludwick, A. E. and Goris, J. E. 1982. Grain protein as an indicator of N sufficiency for winter wheat. Agron. J. 74: 130 133. Grant, C. A. and Flaten, D. N. 1998. Fertilizing for protein content in wheat. Pages 151 168 in D. B. owler, W. E. Geddes, A. M. Johnston, and K. R. Preston, eds. Wheat protein production and marketing. Proc. Wheat Protein Symposium. 9 10 March 1998. University Extension Press, University of Saskatchewan, Saskatoon, SK. Keisling, T. C. and Mullinix, B. 1979. Statistical considerations for evaluating micronutrient tests. Soil. Sci. Soc. Am. J. 43: 1181 1184. Kramer, C. Y. 1956. Extension of multiple range tests to group means with unequal number of replications. Biometrics 12: 309 210. Kramer, T. H. 1979. Environmental and genetic variation for protein in winter wheat (Triticum aestivum L.). Euphytica 28: 209 218. Long, D. S. and Engel, R. E. 1998. Grain protein management using precision framing methods. Pages 169 179 in D. B. Fowler, W. E. Geddes, A. M. Johnston, and K. R. Preston, eds. Wheat protein production and marketing. Proc. Wheat Protein Symposium. 9 10 March 1998. University Extension Press, University of Saskatchewan, Saskatoon, SK. Lawlor, D. W., Kontury, M. and Young, A. T. 1989. Photosynthesis by flag leaves of wheat in relation to protein, ribulose bisphosphate carboxylase activity and nitrogen supply. J. Exp. Bot. 40: 43 52. Lukow, O. M. and Preston, K. R. 1998. Effect of protein content on wheat quality. Pages 48 52 in D. B. Fowler, W. E. Geddes, A. M. Johnston, and K. R. Preston, eds. Wheat protein production and marketing. Proc. Wheat Protein Symposium. 9 10 March 1998. University Extension Press, University of Saskatchewan, Saskatoon, SK. SAS Institute Inc. 1994. JMP statistics and graphic guide. Version 3 of JMP. SAS Institute, Inc., Cary. NC. Selles, F., Zentner, R. P., Read, D. W. L. and Campbell, C. A. 1992. Prediction of fertilizer requirements for spring wheat grown on stubble in southwestern Saskatchewan. Can. J. Soil Sci. 72: 229 241. Smith, E. G., Zentner, R. P., Campbell, C. A., Grant, C. A. and Gehl, D. T. 1998. Economics of fertilization for protein premiums. Pages 128 138 in D. B. Fowler, W. E. Geddes, A. M. Johnston, and K. R. Preston, eds. Wheat protein production and marketing. Proc. Wheat Protein Symposium. March, 9 10, 1998. University Extension Press, University of Saskatchewan, Saskatoon, SK. Steel, G. D. and Torrie, J. H. 1960. Principles and procedures of statistics with special reference to the biological sciences. McGraw-Hill Book Company Inc., New York, NY. Terman, G. L. 1979. Yields and protein content of wheat grain as affected by cultivar, N, and environmental growth factors. Agron. J. 71: 437 440. van Herwaarden, A. F., Fraquhar, G. D., Angus, J. F., Richards, R. A. and Howe, G. N. 1998. Haying-off, the negative grain yield response to nitrogen fertiliser. I. Biomass, grain yield, and water use. Aust. J. Agric. Res. 49: 1067 1081. Virtanen, A. and Peltonen, J. 1996. Post-harvest evaluation of nitrogen sufficiency for small-grain cereals by measuring GPC. J. Agron. Crop Sci. 177: 153 160.