Yield benefit of phosphorus fertilizer for wheat, barley and canola in Alberta

Transcription

1 Yield benefit of phosphorus fertilizer for wheat, barley and canola in Alberta R. H. McKenzie 1, E. Bremer 2, L. Kryzanowski 3, A. B. Middleton 1, E. D. Solberg 3,6, D. Heaney 3,7, G. Coy 4,8 and J. Harapiak 5,9 1 Crop Diversification Division, Alberta Agriculture, Food and Rural Development, Lethbridge, Alberta, Canada T1J 4V6 ( ross.mckenzie@gov.ab.ca); 2 Symbio Ag Consulting, Lethbridge, Alberta, Canada T1K 2B5; 3 Crop Diversification Division, Alberta Agriculture, Food and Rural Development, Edmonton, Alberta, Canada T6H 4P2; 4 Alberta Agriculture, Food and Rural Development, Fairview, Alberta, Canada T0H 1L0; and 5 Westco, Calgary, Alberta, Canada T2C 4M5. Received 24 December 2002, accepted 6 May McKenzie, R. H., Bremer, E., Kryzanowski, L., Middleton, A. B., Solberg, E. D., Heaney, D., Coy, G. and Harapiak, J Yield benefit of phosphorus fertilizer for wheat, barley and canola in Alberta. Can. J. Soil Sci. 83: Crop responsiveness to P fertilizers on the Canadian Prairies has likely declined during the past three to four decades due to regular application of P fertilizer and reduced tillage. Its relationship to extractable soil P as determined by various soil tests may also have changed. The objective of this study was to evaluate five soil test P methods for three major crops across a wide range of soil types and environmental conditions. Small-plot P fertilizer trials were conducted at 154 locations across Alberta from 1991 through At each location, fertilizer responses were determined for one, two, or three crops: barley (Hordeum vulgare L.), spring wheat (Triticum aestivum L.) or canola (Brassica napus L.). Fertilizer treatments consisted of seed-placed monoammonium phosphate at rates of 0, 6.5, 13.1 and 19.6 kg P ha 1. The average increase in seed yield due to application of P fertilizer was 10%, with little difference among crop types. Relative yield increases were significantly greater in Gray soils (Dark Gray Chernozemics, Dark Gray-Gray Luvisols) than in Black (Black Chernozemics) or Brown soils (Brown and Dark Brown Chernozemics). The maximum variation in P fertilizer response accounted for by any soil test P was 27% for barley, 15% for wheat and 7% for canola. The Kelowna method and its derivatives generally provided the best fit with P fertilizer response. Only a modest increase in the proportion of variation that could be accounted for by soil test was achieved by multiple regressions with soil ph, clay or organic matter or by separate analyses of different soil types or years. The probability of a profitable yield response due to P fertilizer application did decline with increasing soil test P. However, profitable yield responses were frequent at all levels of soil test P for the first increment of 6.5 kg P ha 1 and low at all levels of soil test P for the third increment of 6.5 kg P ha 1 (19.6 kg P ha 1 ). The poor relationship of soil test P to fertilizer response was attributed to frequent but variable starter effects of P fertilizer and the infrequent occurrence of highly responsive sites. Key words: Soil testing, Olsen, Bray, Kelowna, fertilizer response functions, Hordeum vulgare, Triticum aestivum, Brassica napus McKenzie, R. H., Bremer, E., Kryzanowski, L., Middleton, A. B., Solberg, E. D., Heaney, D., Coy, G. et Harapiak, J Effets des engrais phosphatés sur le rendement du blé, de l orge et du canola en Alberta. Can. J. Soil Sci. 83: La réactivité des cultures aux engrais phosphatés dans les Prairies canadiennes a sans doute diminué au cours des trente à quarante dernières années en raison de l application régulière de tels engrais et d un moins grand travail du sol. Il se peut aussi que les liens entre cette réactivité et la concentration de P extractible dans le sol, telle que mesurée par divers tests, aient changé eux aussi. L étude devait servir à évaluer cinq tests de dosage du P dans le sol pour trois grandes cultures, selon une grande variété de sols et de facteurs environnementaux. Les auteurs ont utilisé de petites parcelles pour effectuer des essais sur les engrais phosphatés à 154 endroits, en Alberta, de 1991 à À chaque endroit, ils ont déterminé la réaction d une, de deux ou de trois cultures aux engrais, à savoir l orge (Hordeum vulgare L.), le blé de printemps (Triticum aestivum L.) et le canola (Brassica napus L.). Les traitements consistaient en l application de phosphate d ammonium diacide avec la semence à raison de 0, de 6,5, de 13,1 et de 19,6 kg de P par hectare. La hausse moyenne du rendement grainier attribuable à l usage d engrais phosphatés s établit à 10 % et ne varie guère d une culture à l autre. La hausse de rendement est significativement plus importante sur les sols gris (tchernozioms gris foncé, luvisols gris-gris foncé) que sur les sols noirs (tchernozioms noirs) ou bruns (tchernozioms bruns et brun foncé). La plus forte variation attribuable à l engrais phosphaté relevée au moyen d un des tests était de 27 % pour l orge, de 15 % pour le blé et de 7 % pour le canola. La méthode de Kelowna et ses variantes donnent généralement la mesure la plus précise de la réaction aux engrais phosphatés. Les régressions multiples avec le ph, la proportion d argile ou la concentration de matière organique ou l analyse séparée des sols ou des années n augmentent que légèrement la partie de la variation expliquée par les tests. La probabilité d une réaction à la hausse du rendement après application d un engrais phosphaté diminue bel et bien quand on mesure plus précisément la concentration de P dans le sol. Peu importe le test, il est néanmoins fréquent de voir le rendement augmenter pour la première tranche de 6,5 kg de P ajoutée par hectare et de voir cette réaction faiblir quand on atteint la 6 Currently with Sun Mountain Agronomy, Edmonton, Alberta, Canada T7Y 1B8 7 Currently with Norwest Laboratories, Edmonton, Alberta, Canada T6E 0P5. 8 Currently with Aqua Terre Solutions Inc., Calgary, Alberta, Canada T2P 1H4. 9 Currently with Agro/Envro Consulting Services, Calgary, Alberta, Canada T2W 6E2. 431

2 432 CANADIAN JOURNAL OF SOIL SCIENCE locations on Brown soils were always adjacent to a stubble location (same section), while fallow locations on Black or Gray soils were not set up using this approach. Locations had been managed as reduced or conventional tillage systems. At each location, fertilizer responses were determined for one, two, or three crops: barley, spring wheat or canola. Plots from each crop were established in adjacent, but independent, areas at each location. Each trial with one crop at one location will be referred to as a site in the remainder of this paper. Plots for each crop were set up in a randomized complete block design, with six replicates at 80% of the sites and three to five replicates at the remaining sites. Treatments consisted of four rates of commercial monoammonium phosphate: 0, 6.5, 13.1 and 19.6 kg P ha 1. Treatment applications were applied with the seed except at the highest application rate for canola, which, to avoid poor emergence, was split with 13.1 kg P ha 1 seed-placed and 6.5 kg P ha 1 banded prior to seeding. The highest rate of P application was also split for barley and wheat at sites in the Peace region (23 of 42 locations on Gray soils). Plots were seeded using hoe openers at a row spacing of 17.8 cm and a seedfertilizer spread of approximately 1.8 to 2.5 cm. Soil samples were obtained during the previous fall at locations in southern Alberta or just prior to trial establishment at locations in central or northern Alberta. At least two soil samples were obtained at each location (one sample for every three blocks or one per site). Up to 18 soil samples were obtained at some locations (one sample per block per site). Soil samples consisted of about 20 small cores (2 cm) or five large cores (5 cm) and were obtained at depths of 0 15 and cm. All samples were air-dried and ground to pass a 2-mm sieve. Soil test P was measured on all surface (0 15 cm) samples using five different procedures (Table 2). All samples were also analyzed for soil ph (water)(hendershot et al. 1993) and composite samples from each location were analyzed for organic matter based on weight loss at 375 C (16 h) [modification of Karam (1993)] and particle size distribution based on the hydrometer method after dispersal with sodium metaphosphate (Sheldrick and Wang 1993). Sites were cultivated just prior to seeding with a cultivator and a harrow. One or two operations were conducted, depending on the amount of trash. Nitrogen and any other required fertilizers were then banded across the entire site. The bestrated crop varieties were sown in each region. At maturity, total seed weight and moisture content was determined for each plot. Yields are expressed on a dry weight basis. Yield data from each experimental site were subject to an analysis of variance. Fertilizer responses were determined only for experimental sites that had a coefficient of variation of less than 20%. In total, fertilizer responses were detertroisième tranche (19,6 kg de P par hectare). La piètre relation entre le dosage du P dans le sol et la réaction aux engrais résulte de l effet de démarrage fréquent mais variable de l engrais phosphaté et du petit nombre de sites se caractérisant par une vive réaction à l amendement. Mots clés: Analyse de sol, Olsen, Bray, Kelowna, fonctions de réaction aux engrais, Hordeum vulgare, Triticum aestivum, Brassica napus Phosphorus fertilizers are applied annually to almost all cereal and oilseed crops grown on the Canadian prairies. Their benefit to crop yield was first demonstrated in studies conducted from 1928 to 1930 (Mitchell 1932) and their use expanded rapidly from the early 1940s until the late 1960s (Mitchell 1946; Doyle and Cowell 1993). Since 1975, the import of P in fertilizers to the three Prairie Provinces has been approximately equal to the export of P in grain (Doyle and Cowell 1993). The widespread use of P fertilizers has likely contributed to a reduced yield benefit of P fertilizer: prior to 1970, fertilizer trials in Saskatchewan showed an average yield increase for wheat of 26% (874 trials), but after 1970, the average yield increase for wheat was only 11% (252 trials) (Cowell and Doyle 1993). The widespread adoption of reduced tillage systems has also increased P availability from residual P fertilizer (Black 1982). A number of soil test methods have been used for P fertilizer recommendations on the Canadian prairies. Mitchell (1932) observed a relatively close relationship of the P response of wheat to the measurement of available soil P using a modified Truog extractant (K 2 SO 4 /H 2 SO 4 ). Extensive field-testing in the 1950s and 1960s led to the establishment of the Olsen method (Olsen et al. 1954) as the basis for P fertilizer recommendations in Saskatchewan (Cowell and Doyle 1993). In Alberta, early work led to the establishment of the Miller-Axley method as the basis for P fertilizer recommendations (Robertson 1962). In the past decade, laboratories that now do most of the soil testing in western Canada have adopted modifications of the Kelowna method (Qian et al. 1994; Ashworth and Mrazek 1995). The main advantages of the Kelowna-type tests are their analytical convenience and applicability to both calcareous and non-calcareous soils (Van Lierop 1988). Uncertainty in the prediction of P fertilizer requirements exists due to the changing methods of estimating available P and the probable change in the responsiveness of crops to P fertilizer. The objective of this study was to evaluate the ability of the most widely used methods of estimating available soil P for the prediction of optimum P fertilizer rates for barley, wheat and canola. MATERIALS AND METHODS Small plot fertilizer trials were conducted at 154 locations across Alberta from 1991 through 1993 (Table 1). The locations ranged from Foremost (49 48 N, W) in southeastern Alberta to Manning (56 92 N, W) in northwestern Alberta. Twenty-eight percent of locations were on Brown and Dark Brown Chernozemics (Brown soils), 45% were on Black Chernozemics (Black soils) and 27% were on Dark Gray Chernozemics and Dark Gray- Gray Luvisols (Gray soils). About 80% of the locations were on stubble land and 20% were on fallow land. Fallow

3 MCKENZIE ET AL. YIELD BENEFIT OF P FERTILIZER IN ALBERTA 433 mined at 143 barley sites, 141 wheat sites and 108 canola sites (Table 1). Fertilizer response coefficients were determined for each site by fitting the data to the model described in Fig. 1. The model splits the fertilizer response into three phases: (1) a deficient phase where nutrient supply is limiting for maximum crop yield and crop yield increases with increasing fertilizer rate following an exponential function (modification of Mitscherlich equation), (2) a sufficient phase where nutrient supply is non-limiting to crop growth and crop yields are unaffected by fertilizer rate and (3) an excessive phase where crop yields decline with increasing fertilizer rate due to seed toxicity, induced nutrient imbalance or other negative impacts of high rates of fertilizer application. Although in theory the coefficients in the model could be determined using non-linear regression techniques, an analytical approach was used because only four rates of fertilizer were included in this study and coefficients could be determined more reliably and quickly by iterative fitting. Fertilizer rates were considered to be sufficient if crop yields were more than 96% of the highest mean yield. Most sites had two or more rates in this range, and the average of all of the yields at these rates was used to estimate maximum yield (Y max ). If only one rate of fertilizer had a yield of more than 96% of the highest yield, this yield was used to estimate Y max except at two barley sites, where yields were still increasing linearly at the highest rate of fertilizer application and Y max was estimated at 110% of the highest mean yield. The Mitscherlich equation requires estimates of maximum potential yield (M), fertilizer equivalence of soil nutrient supply (b), and curvature (c) of the response of yield to increasing fertilizer rate (x) (Holford et al. 1985; Cerrato and Blackmer 1990): Y = M (1 e c(b + x) ) The equation was modified to force the estimate of Y to reach Y max at 98% of M: M = Y max /0.98 The value of b was calculated at x = 0: b = LN(1 Y P = 0 /M)/c Curvature (c) was estimated from crop yields at fertilizer rates in the deficient range: c = LN(1 Y x /M)/(b + x) Several iterations were required to estimate c and b. At sites with yield losses of greater than 4% at high rates of fertilizer application, a yield reduction coefficient (d) was calculated based on the yield at the highest rate of fertilizer application: d = (Y max Y P = 19.6 )/(19.6 x s ) 2 Table 1. Number of experimental sites with range of soil properties and mean monthly temperature and precipitation Soil type Variable Brown Black Gray All Number of locations Total Stubble Fallow Number of sites per crop Barley Wheat Canola Soil ph (0 15 cm) Median Minimum Maximum Clay content (%, 0 15 cm) Median Minimum Maximum Soil organic matter (%, 0 15 cm) Median Minimum Maximum Soil test P (mg kg 1, 0 15 cm, KEL method) Median Minimum Maximum Average monthly temperature z ( C) May June July Aug Monthly precipitation z (mm) May June July Aug Total z Means from eight Environment Canada weather stations located across Alberta. where x s is the highest safe rate of fertilizer application (i.e., highest rate of fertilizer in the sufficient range). Economic optimum rates of fertilization (P opt ) were estimated using fertilizer-to-grain price ratios (F) of 15 for barley, 10 for wheat and 6 for canola: P opt = {LN(F) LN( M e bc c)}/ c Statistical comparisons of fertilizer response among crop types or other variables were determined by ANOVA procedures using each site or location as one replicate. Linear, non-linear and multiple regression procedures were used to evaluate the relationships of measured soil properties to fertilizer response. RESULTS AND DISCUSSION Fertilizer Response Fertilizer responses were completely in the deficient range at 18% of all sites (e.g., Loc#140, Fig. 2) and completely in

4 434 CANADIAN JOURNAL OF SOIL SCIENCE Table 2. Soil test P methods Soil test Acronym Extractant Soil to extractant ratio Shaking time (min) Reference Kelowna KEL 0.15 M NH 4 F M Van Lierop (1988) CH 3 COOH Modified Kelowna #1 MK M NH 4 F Qian et al. (1994) 0.25 M CH 3 COONH M CH 3 COOH Modified Kelowna #2 MK M NH 4 F Ashworth and 1.0 M CH 3 COONH 4 Mrazek (1995) 0.5 M CH 3 COOH Miller-Axley MA 0.03 M NH 4 F 5 10 Miller and Axley M H 2 SO 4 (1956) (modified) Olsen OLS 0.5 M NaHCO 3, ph Olsen et al. (1954) Fig. 1. Fertilizer response function used at all experimental sites. Only the appropriate portion of the response function was used at individual sites. For details, see text. the sufficient to excessive range at 17% of all sites (e.g., Loc#54, Fig. 2). Most sites (65%) were in both the deficient and sufficient ranges (e.g., Loc#19, 33, and 42, Fig. 2). Statistically significant reductions in yield at high rates of fertilizer application were observed at 2.5% of all sites, while yield declines of more than 4% were observed at 12% of all sites. More than 80% of predicted yield increases using the model described in Fig. 1 were within 125 kg ha 1 of actual values, with a coefficient of determination (R 2 ) between predicted and actual yield of greater than 0.90 for all crops. However, not too much significance should be given to this close fit because most functions can provide a good fit to fertilizer response data, particularly if there are relatively few rates of fertilizer application (Cerrato and Blackmer 1990). The major benefit of the model was that the full range of P fertilizer response could be quickly and precisely described. One advantage suggested for the use of the Mitscherlich function is that it provides coefficients that are meaningful and comparable among sites: b is a measure of soil nutrient supply expressed in fertilizer equivalents and c is a measure of the curvature of the fertilizer response function (Holford et al. 1985). Although in theory these coefficients should be independent, a close relationship was observed between b

5 MCKENZIE ET AL. YIELD BENEFIT OF P FERTILIZER IN ALBERTA 435 Fig. 2. Illustration of fitted functions at sites that were completely in the deficient range (#140), completely in the sufficient range (#54), or in the deficient to sufficient (#19, 42) or excessive (#33) ranges. and c in this study: c = 0.64/SQRT(b) (R 2 = 0.67, n = 326). This relationship occurred because the fertilizer response at many sites had a high curvature (e.g., Loc#42, Fig. 2), which occurs when low rates of fertilizer are sufficient to increase yields to values that are close to maximum. Other studies have shown that P nutrition early in the growing season is critical for good crop growth and that low rates of P fertilizer in close proximity to the young plant can provide large improvements in crop growth, even at high levels of P fertility (Grant et al. 2001). This starter effect of P fertilizer likely contributed to the wide variation of curvature observed in this study (0.03 to 0.36), which invalidates direct comparisons of soil nutrient supply estimates (b) among sites. Two-thirds of the cereal sites and just under half of the canola sites had a significant (P < 0.05, LSD) yield increase due to P application (Table 3). The frequency of sites with a significant yield increase was similar to that reported by Mahli et al. (1993) for central Alberta, where 54% of barley trials (n = 80) and 48% of canola trials (n = 48) had a significant yield increase due to application of P fertilizer. The average (median) increase in yield due to application of P fertilizer was 10% (Table 3). In comparison, Cowell and Doyle (1993) reported that P fertilizer increased wheat yield by an average of 11% in 252 trials conducted across Saskatchewan between 1970 and They reported that the P response in recent trials was considerably less than obtained in trials conducted between 1939 and 1969, when P fertilizer increased wheat yield by an average of 26% in 874 trials. This was attributed primarily to the widespread Table 3. Effect of crop type on P fertilizer response (all locations) Frequency Yield gain P x opt (Y P0 <Y max ) z (Y max Y P0 ) Crop (% of sites) (% of Y P0 ) (kg P ha 1 ) All locations Barley (n = 143) Wheat (n = 141) Canola (n = 108) All locations with both barley and wheat present (n = 132) Barley Wheat P (paired t-test) * y NS R All locations with both barley and canola present (n = 104) Barley Canola P (paired t-test) NS NS R All locations with both wheat and canola present (n = 103) Wheat Canola P (paired t-test) * NS R z Sites with significant (P < 0.05, LSD) yield gain, Y P0 = yield with no P fertilizer, Y max = maximum yield. Probability: NS = Not significant (P > 0.1), *P < 0.1. x P opt = economic optimum rate of P fertilizer addition. adoption of P fertilizer and reduced tillage, which gradually builds up plant-available residual P fertilizer in surface soils (Black 1982; Cowell and Doyle 1993). Differences in average P fertilizer response were small among crop types: the only significant difference was a

6 436 CANADIAN JOURNAL OF SOIL SCIENCE Table 4. Effect of soil type and year on P fertilizer response Frequency (% of sites with Y P0 <Y max ) z Yield gain (% of Y P0 ) Crop Soil type All All Barley Brown Black Gray All Wheat Brown Black Gray All Canola Brown Black Gray All Summary of ANOVA Barley Wheat Canola Prob. Soil type * y NS * Year NS NS * SxY NS NS ** z Sites with significant (P < 0.05, LSD) yield gain, Y P0 = yield with no P fertilizer, Y max =maximum yield. y NS = not significant (P > 0.05); *P < 0.05; **P < slightly smaller yield gain for wheat than for barley or canola (Table 3). Grant and Bailey (1993) suggested that canola might be able to achieve maximum yields at lower rates of P than wheat or barley because it is very effective at acquiring soil and fertilizer P. In this study, significant yield increases due to P application were less frequent for canola than cereal crops, although this was at least partly due to the higher variability in canola yield measurements (average coefficient of variation of 8.8% for canola vs. 6.4% of cereals). Cowell and Doyle (1993) reported a slightly smaller yield benefit of P fertilizer for canola (8%, n = 78) than wheat (11%, n = 252) in trials conducted across Saskatchewan between 1970 and The small differences in optimum fertilizer rates among crop types indicate that one fertilizer response function might be sufficient for all three crops. However, linear regression accounted for less than 10% of the variation in optimum fertilizer rates between canola and either cereal crop and only 18% of the variation in optimum fertilizer rate between barley and wheat. Soil type and year significantly affected P fertilizer response, depending on crop type (Table 4). The yield benefit of P fertilizer, expressed as percent of unfertilized control, was significantly greater for barley and canola on Gray than on Black or Brown soils. However, a significant interaction with year occurred for canola: the frequency and magnitude of P fertilizer benefit were similar among soil types in 1993, but were greatest on Gray soils in 1991 and The effect of soil type was similar in all years for barley, but the frequency and magnitude of the P fertilizer benefit on Gray soils in 1991 were lower than in 1992 or 1993 while the frequency of fertilizer benefits was much higher in 1993 than 1991 or 1992 on Brown soils. This contrasting response to year and soil type accounts for part of the poor correlation observed in fertilizer response between crop types (Table 3). One factor that contributes to differences in P fertilizer response among crop types is P acquisition strategy. Canola roots lower the ph within the rhizosphere more effectively than cereal roots, allowing canola to deplete acid-soluble P fractions more effectively than cereals (Grant and Bailey 1993; McKenzie et al. 1995). This strategy is likely to be most effective in calcareous soils and may partially account for the less-frequent and smaller response of canola in Brown and Black soils. A second factor that may contribute to differences in crop response to P fertilizer is the starter effect of P fertilizer. Starter effects of P fertilizer are frequent in cool soils, but depend on crop type, soil conditions and fertilizer placement (Zentner et al. 1993; Grant et al. 2001). The importance of starter fertilizer in cool soils may account for the greater P fertilizer response in Gray soils. Average monthly air temperatures in May and June were 1.3 C cooler in regions with Gray soils than Brown soils during this study (data not presented). Differences in soil temperature also depend on soil moisture (generally wetter in Gray soils) and date of seeding (2 to 3 wk later in Gray soils). The impact of differences in early season growth on final yield, however, depends strongly on conditions during later growth, particularly precipitation. Mitchell (1946) indicated that wheat responses to P fertilizer were least when a hot, dry summer followed a cool spring, which was attributed to early-season depletion of soil moisture that was critical for grain yield formation. This response is consistent with the less-frequent yield benefits of P fertilizer for cereals in all soil types in 1991, which was characterized by warm, dry conditions during the summer (Table 1). Canola has more indeterminate growth than cereals and differences in early growth often do not translate into differences in seed yield due to compensatory growth later in the growing season. Favourable moisture conditions throughout the growing season may increase the ability of canola to com-

7 MCKENZIE ET AL. YIELD BENEFIT OF P FERTILIZER IN ALBERTA 437 pensate for slower early growth. This may account for the weaker P response by canola in 1993, the year with the most precipitation, in contrast to the similar or stronger P response by cereals in 1993 than in previous years (Tables 1 and 4). Crop type also influenced the impact of summer fallow on P fertilizer response. The frequency of significant yield increases, relative yield gain and optimum fertilizer rate were all significantly higher on fallow than on stubble sites for wheat, but were unaffected by previous land management for barley and canola (Table 5). Similar differences in the P fertilizer response between fallow-grown and stubblegrown wheat were reported in Saskatchewan studies (Cowell and Doyle 1993; Zentner et al. 1993). Measurements of available P for the sites included in Table 5 were almost identical for fallow and stubble sites (18 vs. 16 mg P kg 1, respectively, based on the KEL method). While the cause for this difference in response for wheat is not known, wheat might respond more strongly to P fertilizer on fallow due to depletion of organic P during the fallow period, reduced mycorrhizal infection or increased yield potential. Relation of P Fertilizer Response to Soil Test P All of the soil test P methods in this study were highly correlated, although differences were still evident (Table 6). The two derivatives of the KEL method (MK1 and MK2) extracted slightly less P than the KEL method, but remained very highly correlated with the KEL method and each other. These methods had a weak negative correlation with soil ph, a weak positive correlation with clay content, and no correlation with soil organic matter. Available P determined using the OLS method had a similar negative correlation with soil ph as the KEL method, but was more strongly correlated with clay concentration. In contrast, the MA method had a strong negative correlation with soil ph and no correlation with clay or organic matter. This strong negative correlation can be attributed to increasing neutralization of acid and possibly increasing precipitation of fluoride as CaF 2 in the extracting solution with increasing soil ph (Van Lierop 1988). The relationship between the MA and KEL methods could be significantly improved by including ph in the regression (MA = KEL 5.1 ph, R 2 = 0.74). None of the soil test methods were very effective for predicting P fertilizer response, particularly for canola and wheat (Table 7). When all sites were included in the correlation, the Kelowna-type methods provided the best fit with P fertilizer yield gains, but still only accounted for up to 27% of the variability in yield gain for barley, 15% of the variability in yield gain for wheat, and 7% of the variability in yield gain for canola. Correlations based on individual soil zones show improved fits in some soil zones, but not others, independently of soil test method for the most part (Table 7). Coefficients of determination of soil test methods with optimum fertilizer rates were very similar to those with yield gain (data not presented). In all cases, the range in yield gain or optimum P fertilizer rate was wide at all values of soil test P, but tended to decrease with increasing soil test P concentration (e.g., Fig. 3). Table 5. Effect of cropping practice in previous year on P fertilizer response z Yield gain (Y max Y P0 ) P opt x Crop Previous year (% of sites) (kg ha 1 ) (% of Y P0 ) (kg P ha 1 ) Barley Summer fallow a x 8a 8a (n = 17) Cereal crop a 11a 8a Wheat Summer fallow a 12a 11a (n = 17) Cereal crop a 8b 6b Canola Summer fallow a 12a 9a (n = 16) Cereal crop a 9a 7a z Only locations that had adjacent sites on summer fallow and cereal stubble were included. All locations were on Brown soils. y Sites with significant (P < 0.05, LSD) yield gain, Y P0 = yield with no P fertilizer, Y max =maximum yield. x P opt = economic optimum rate of P fertilizer addition. a,b Values followed by the same letter within a column are not significantly different (P < 0.05, protected LSD). Multiple regression techniques were used to determine if other soil properties could improve the correlation of soil test P with P fertilizer response (Table 8). The only variable that significantly improved the correlation for cereal crops was soil ph. For barley, yield gain due to application of P fertilizer declined by an average of 165 kg ha 1 per unit increase in soil ph while optimal rates of P fertilizer application declined by an average of 2.4 kg P ha 1 per unit increase in soil ph. Pair-wise comparisons showed that barley fertilizer response was significantly greater at sites when soil ph was <6 than 6; similar comparisons were not significant if the critical ph was increased above 6 for barley or for wheat and canola at any set ph (data not presented). Either direct or indirect relationships may account for the increased response to P fertilizer of barley in acid soils (<ph 6). The only variable that significantly improved the correlation for canola was soil organic matter. The increase in yield response and fertilizer requirement of canola with increasing organic matter is likely due to the correlation of organic matter with temperature and moisture: sites with high organic matter are located in the cooler and moister Gray and Black soil zones (Table 1). Phosphorus fertilizer response was not significantly affected by soil clay content (Table 8), but tended to increase with increasing clay content, consistent with earlier studies (Mitchell 1946; Cowell and Doyle 1993). Evaluation of the relationships of P fertilizer response to other soil properties within soil zones generally confirmed the conclusions based on all sites. Overall, improvements in R 2 were modest due to inclusion of other soil properties (Tables 7 and 8). Another factor that has frequently been used to improve fertilizer prediction is potential crop yield. Crops with a high yield potential due to good growing conditions (e.g., good moisture availability) have a greater demand for nutrients and may obtain a greater economic benefit from higher fertilizer rates than crops with a low yield potential (Barber 1973). However, in this study, no correlation was observed between maximum yields and yield gain or optimal rate of fertilizer application (R , any crop). The lack of any correlation indicates that P fertilizer recommendations may

8 438 CANADIAN JOURNAL OF SOIL SCIENCE Table 6. Correlation matrix of soil test P methods and soil properties (0 15 cm) be developed independently of potential yield for crops in this region. The poor fit between soil test P and yield benefit of P fertilizer application can be attributed to several factors: (1) The weak response to P fertilizer and inherent variability in measurement limit the proportion of variability that could be accounted for. Attainable values of R 2 are reduced by random error: Maximum R 2 = (total SS pure error SS)/total SS (Draper and Smith 1981). Based on measurement errors at each site, the maximum R 2 for yield gain in this study was 0.77 for barley, 0.50 for wheat and 0.59 for canola. Higher values of R 2 are potentially attainable only with a greater yield response or a lower measurement error. (2) The influence of factors other than available P on fertilizer response also limited the proportion of variability that could be accounted for. As noted previously, factors that affected P fertilizer response include crop type, year (weather variables), soil type and prior crop history. Segregating out these factors (e.g., year and soil type) provided some improvement in goodness-of-fit. Further improvement would require a much better understanding and measurement or control of relevant factors. Coefficient of correlation (r) Variable z Slope y KEL MK1 MK2 MA OLS ph CLAY MK *** MK *** 0.95*** MA *** 0.82*** 0.77*** OLS *** 0.93*** 0.89*** 0.77*** ph ND x 0.19* 0.21* *** 0.18* CLAY ND 0.23** 0.21* 0.21** *** 0.11 OM ND * 0.17* z See Table 2 for description of variables. y Slope between variable and soil test P determined using the Kelowna method. The intercept was set to zero. x Not determined. *, **, *** P < 0.05, P < 0.01 and P < 0.001, respectively Table 7. Coefficients of determination (R 2 ) between soil test P method and P fertilizer response z R 2 {Ln(STP) vs. Y max Y P0 (kg ha 1 )} Crop Soil type KEL MK1 MK2 MA OLS Barley All Brown 0.17 NS NS NS NS Black Gray NS NS Wheat All NS 0.08 Brown NS NS NS Black NS 0.17 Gray NS y NS NS NS NS Canola All NS NS NS Brown NS NS NS NS NS Black NS 0.24 NS Gray NS NS NS NS NS z STP = soil test P (0 15 cm), Y P0 = yield with no P fertilizer, Y max = maximum yield. z Not significant (P > 0.01). (3) The soil test methods may have provided an unreliable estimate of available P because crops were able to acquire P that was not measured by the soil test method or because crops were unable to acquire P that was measured by the soil test method. For example, soil test P measurements were extremely low for the MA method when soil ph > 7.5, but crop response (and other soil test methods) showed that available P was often high. However, the ability to detect problems with soil test methods in this study was limited due to the previous two reasons. Graphical methods are often used to determine critical soil test levels to separate soils that are likely to respond to fertilizer application from those that will not (Dahnke and Olson 1990). Based on a number of studies, Olsen and Sommers (1982) concluded that P fertilizer responses were unlikely if soil test P (OLS) was greater than 10 mg kg 1. Soon (1990) reported critical levels of soil test P (OLS) of 11 to 14 mg kg 1 based on greenhouse and field trials with barley and canola in northwestern Alberta. Black (1982) and Jackson et al. (1997) reported critical levels of soil test P (OLS) of 15 to 16 mg kg 1 for spring wheat in Montana. Similar determinations in this study were difficult due to the frequent responsiveness at high soil test P levels and non-responsiveness at low soil test levels (Fig. 3). This was also a problem in previous studies, as indicated by low relative yields at critical soil test P levels and the number of outliers (Soon 1990; Jackson et al. 1997). This occurs because critical soil test P levels are likely much higher during early than later crop growth due to the small root system, low soil P solubility, slow P diffusion and possibly greater physiological demand for P during this period (Grant et al. 2001). A more useful method of presenting fertilizer recommendations when factors other than fertility greatly influence crop response is based on probability diagrams (Fits 1955). Using the response functions determined for each site and fertilizer-to-grain cost ratios, the probability (% of sites) of a profitable increase in yield as a function of soil test and fertilizer rate can be determined (Fig. 4). The probability of a profitable increase in yield due to application of the first 6.5 kg P ha 1 was high for all crops and soil test levels, declining from close to 100% when soil test P (KEL) was less than 10 mg P kg 1 (very low) to about 60%

9 MCKENZIE ET AL. YIELD BENEFIT OF P FERTILIZER IN ALBERTA 439 Fig. 3. Yield benefit of P fertilizer for wheat as a function of soil test P (0 15 cm, KEL method). The bold line and equation are based on a regression of all sites; the other lines are based on regressions from each soil type and year. Table 8. Multiple regression coefficients of soil properties (0 15 cm) with P fertilizer response Variables Crop Constant Ln(KEL) ph Clay (%) OM (%) R 2 Coefficients with yield gain (Y max Y P0, kg ha 1 ) Barley 2461*** 376*** 165*** 1.31NS 15.9NS 0.36 Wheat 1112*** 162*** 60* 1.46NS 2.7NS 0.16 Canola 450NS 78* 21NS 0.72NS 17.6* 0.10 Coefficients with optimum P rate (kg P ha 1 ) Barley 40.5*** 6.5*** 2.4** 0.03NS 0.37NS 0.31 Wheat 29.4*** 4.0*** 1.4* 0.04NS 0.14NS 0.12 Canola 18.7NS 2.4NS 0.8NS 0.00NS 0.57* 0.08 *, **, *** P < 0.05, P < 0.01, and P < 0.001, respectively; NS, not significant. when soil test P was more than 30 mg P kg 1 (high). Crop type and soil zone had little influence on the probability of a profitable increase with the first increment of fertilizer addition. With the second increment of 6.5 kg P ha 1, a wider variation in the probability was observed, particularly at the lowest level of soil test P. The probability of a profitable P fertilizer response at the lowest soil test level was least for barley and canola in Brown soils. The probability of a profitable increase in yield with the third increment of 6.5 kg P ha 1 was low for all soil test levels and crops. Based on calculated fertilizer cost and crop value at each site, maximum profits were achieved if sites were fertilized at rates that provided an approximate 40% probability of a profitable increase. At this level, rates of required P fertilizer ranged from about 10 kg P ha 1 at high soil test P levels to about 20 kg P ha 1 at low soil test P levels (Fig. 4). CONCLUSIONS The average yield response to P fertilizer at the 154 locations included in this study was 10%. This increase was much smaller than reported in early studies on the Canadian Prairies, likely due to a gradual increase in soil P fertility caused by the regular application of P fertilizer. Despite this weak response, the application of up to 10 to 20 kg P ha 1 often provided a positive net return. The probability of a profitable yield benefit declined with increasing fertilizer rate or soil test P level. Soil test P methods were highly correlated with each other, but were not highly correlated with yield gains or optimum P fertilizer rates due to the weak response of yield to P fertilizer and the importance of factors other than soil fertility for P fertilizer response. The Kelowna method and its derivatives generally provided the best fit with P fertilizer response. ACKNOWLEDGEMENTS The authors gratefully acknowledge the field staff of Alberta Agriculture, Food and Rural Development and Westco for assistance in conducting the field trials. We would also like to thank the Agri-Food Laboratory Branch, Alberta Agriculture, Food and Rural Development, for soil analyses. Funding for this project was provided by Alberta Agriculture Research Institute, Western Grains Foundation,

10 440 CANADIAN JOURNAL OF SOIL SCIENCE Fig. 4. Probability of an economic yield increase with increasing rate of P fertilizer and soil test P level. Probabilities are based on the percentage of sites that had an economic increase in crop yield due to application of the first (a), second (b) or third (c) increment of 6.5 kg fertilizer P ha 1 at very low (<10 mg P kg 1 ), low (10 to 20 mg P kg 1 ), medium (20 to 30 mg P kg 1 ) and high (>30 mg P kg 1 ) levels of soil test P (KEL method). Only probabilities based on more than 4 sites are included. Alberta Canola Producers Commission, Westco, Agrium and Potash and Phosphate Institute of Canada. Ashworth, J. and Mrazek, K Modified Kelowna test for available phosphorus and potassium in soil. Commun. Soil Sci. Plant Anal. 26: Barber, S. A The changing philosophy of soil test interpretations. Pages in L. M. Walsh and J. D. Beaton, eds. Soil testing and plant analysis. Rev. ed. SSSA, Madison, WI. Black, A. L Long-term N-P fertilizer and climate influences on morphology and yield components of spring wheat. Agron. J. 74: Cerrato, M. E. and Blackmer, A. M Comparison of models for describing corn yield response to nitrogen fertilizer. Agron. J. 82: Cowell, L. E. and Doyle, P. J The changing fertility of prairie soils. Pages in D. A. Rennie, C. A. Campbell, and T. L. Roberts, eds. Impact of macronutrients on crop responses and environmental sustainability on the Canadian prairies. Canadian Society of Soil Science, Ottawa, ON. Dahnke, W. C. and Olson, R. A Soil test correlation, calibration, and recommendation. Pages in R. L. Westerman, ed. Soil testing and plant analysis, 3rd ed.-sssa Book Series, no. 3. SSSA, Madison, WI.

11 MCKENZIE ET AL. YIELD BENEFIT OF P FERTILIZER IN ALBERTA 441 Doyle, P. J. and Cowell, L. E Balance of nutrient inputs (fertilizers) and exports (grain) in Alberta, Manitoba and Saskatchewan. Pages 1 25 in D. A. Rennie, C. A. Campbell, and T. L. Roberts, eds. Impact of macronutrients on crop responses and environmental sustainability on the Canadian prairies. Canadian Society of Soil Science, Ottawa, ON. Draper, N. R. and Smith, H Applied regression analysis. 2nd ed. John Wiley & Sons, New York, NY. 709 pp. Fits, J. W Using soil tests to predict a probable response from fertilizer application. Better Crops Plant Food 39(3): Grant, C. A. and Bailey, L. D Fertility management in canola production. Can. J. Plant Sci. 73: Grant, C. A., Flaten, D. N., Tomasiewicz, D. J. and Sheppard, S. C The importance of early season phosphorus nutrition. Can. J. Plant Sci. 81: Hendershot, W. H., Lalande, H. and Duquette, M Soil reaction and exchangeable acidity. Pages in M. R. Carter, ed. Soil sampling and methods of analysis. Canadian Society of Soil Science, Lewis Publishers, CRC Press, Boca Raton, FL. Holford, I. C. R., Morgan, J. M., Bradley, J. and Cullis, B. R Yield responsiveness and response curvature as essential criteria for the evaluation and calibration of soil phosphate tests for wheat. Aust. J. Soil Res. 23: Jackson, G. D., Kushnak, G. D., Carlson, G. R. and Wichman, D. M Correlation of the Olsen phosphorus soil test: spring wheat response. Commun. Soil Sci. Plant Anal. 28: Karam, A Chemical properties of organic soils. Pages in M. R. Carter, ed. Soil sampling and methods of analysis. Canadian Society of Soil Science, Lewis Publishers, CRC Press, Boca Raton, FL Malhi, S. S., Nyborg, M., Penney, D. C., Kryzanowski, L., Robertson, J. A. and Walker, D. R Yield response of barley and rapeseed to P fertilizer: influence of soil test P level and method of placement. Commun. Soil Sci. Plant Anal. 24: McKenzie, R. H., Dormaar, J. F., Schaalje, G. B. and Stewart, J. W. B Chemical and biochemical changes in the rhizospheres of wheat and canola. Can. J. Soil Sci. 75: Miller, J. R. and Axley, J. H Correlation of chemical soil tests for available phosphorus with crop response, including a proposed method. Soil Sci. 82: Mitchell, J A preliminary investigation on determining the available phosphorus in Saskatchewan soils. Sci. Agric. 12: Mitchell, J The effect of phosphatic fertilizers on summerfallow wheat crops in certain areas of Saskatchewan. Sci. Agric. 26: Olsen, S. R. and Sommers, L. E Phosphorus. Pages in Methods of soil analysis, Part 2. Chemical and microbiological properties. Agronomy Monograph no. 9. 2nd ed. ASA- SSSA, Madison, WI. Olsen, S. R., Cole, C. V., Watanabe, F. S. and Dean, L. A Estimation of available phosphorus in soils by extraction with sodium bicarbonate. US Department of Agriculture, Washington, DC. Circ Qian, P., Schoenau, J. J. and Karamanos, R. E Simultaneous extraction of available P and K with a new soil test: A modification of Kelowna extraction. Commun. Soil Sci. Plant Anal. 25: Robertson, J. A Comparison of an acid and an alkaline extracting solution for measuring available phosphorus in Alberta soils. Can. J. Soil Sci. 42: Sheldrick, B. H. and Wang, C Particle size distribution. Pages in M. R. Carter, ed. Soil sampling and methods of analysis. Canadian Society of Soil Science, Lewis Publishers, CRC Press, Boca Raton, FL. Soon, Y. K Comparison of parameters of soil phosphate availability for the northwestern Canadian Prairie. Can. J. Soil Sc. 70: Van Lierop, W Determination of available phosphorus in acid and calcareous soils with the Kelowna multiple-element extractant. Soil Sci. 146: Zentner, R. P., Campbell, C. A. and Selles, F Build-up in soil available P and yield response of spring wheat to seed-placed P in a 24-year study in the Brown Soil zone. Can. J. Soil Sci. 73:

12