Phosphorus use efficiency and long-term trends in soil available phosphorus in wheat production systems with and without nitrogen fertilizer

Transcription

1 Phosphorus use efficiency and long-term trends in soil available phosphorus in wheat production systems with and without nitrogen fertilizer F. Selles 1, C. A. Campbell 2, R. P. Zentner 3, D. Curtin 4, D. C. James 3, and P. Basnyat 3 1 Agriculture and Agri-Food Canada, P.O. Box 1000A, Brandon, Manitoba, Canada R7A 5Y3; 2 Agriculture and Agri-Food Canada, 960 Carling Ave. Ottawa, Ontario, Canada K1A 0C6; 3 Agriculture and Agri-Food Canada, P.O. Box 1030, Swift Current, Saskatchewan, Canada S9H 3X2; and 4 New Zealand Institute for Plant and Food Research, Private Bag 4704, Christchurch, New Zealand. Received 22 April 2010, accepted 9 November Selles, F., Campbell, C. A., Zentner, R. P., Curtin, D., James, D. C. and Basnyat, P Phosphorus use efficiency and long-term trends in soil available phosphorus in wheat production systems with and without nitrogen fertilizer. Can. J. Soil Sci. 91:3952. Efficient use of phosphorus (P) in crop production is important for economic and environmental reasons, and to prolong the life of a limited resource. Short-term studies often show low recovery of fertilizer P, but P use efficiency may be underestimated because the value of residual P in the soil is ignored. Our objective was to determine fertilizer P use efficiency in two wheat production systems [continuous wheat (CW) and a 3-yr rotation of summer fallow-wheat-wheat (FWW)] using data from a 39-yr study ( ) at Swift Current, SK. Each rotation received either P only (P) or nitrogen plus P (NP) fertilizer. Annual grain P removal was monitored (all straw was returned to the soil) and changes in soil available P (0- to 15-cm layer) were measured by the Olsen bicarbonate method. In 1993, subplots which received no additional P were established to evaluate the residual effect of P fertilizer applied in the preceding 27 yr. Where P was applied each year, grain P removal averaged 54 to 78% of fertilizer P, with values as high as 65 to 109% in 1994 to 2005, the period of lowest water deficit. The P-only treatments removed 13% less P in grain, on average, than NP treatments. In the P-nly systems, Olsen P content increased linearly with time, but in the NP systems it reached a maximum after 2022 yr and then stabilized. The cumulative P balance (fertilizer P minus P removed in grain) accounted for 60% of the variability in Olsen P accumulation over the course of the experiment. In CW, Olsen P content increased by 0.15 kg ha 1 for each kg ha 1 of P added in excess of crop removal. The rate of Olsen P accumulation was greater (0.20 kg ha 1 for each kg ha 1 of excess fertilizer P) in the FWW rotation possibly due to P mineralization during the summer fallow year. When P was withheld between 1994 and 2005, total grain production in the CW rotation was reduced slightly (by 10%), but there was no significant effect on FWW. Crop P removal ( ) where P was withheld in the final 12 yr was equivalent to 105 and 90% of fertilizer P added to the NP and P-only systems, respectively. We concluded that residual P in prairie soils is retained in forms that are available to plants; wheat crops may therefore recover close to 100% of applied fertilizer P given sufficient time. Key words: Crop production, P balance, Olsen P, fertilizer-p use efficiency, Brown Chernozem Selles, F., Campbell, C. A., Zentner, R. P., Curtin, D., James, D. C. et Basnyat, P Efficacite de l utilisation du phosphore et tendance à long terme de la concentration du phosphore disponible dans le sol pour les syste` mes de production du ble avec ou sans apport d azote. Can. J. Soil Sci. 91:3952. Il importe de savoir avec quelle efficacite les cultures utilisent le phosphore (P), tant pour des motifs économiques et environnementaux que pour prolonger la dure e d une ressource limite e. Les e tudes à court terme indiquent souvent que les plantes assimilent lentement le P des engrais, mais il se peut qu on sous-estime l efficacite avec laquelle le P est utilise parce qu on ne glige l importance du P re siduel dans le sol. Les auteurs ont voulu e tablir l efficacite de l assimilation du P par deux systèmes de production du ble [monoculture (MC) et assolement de trois ans jache` re-ble -ble (JBB)], à partir des données d une e tude de 39 ans ( ) re alise e à Swift Current (Saskatchewan). Chaque syste` me a reçu un engrais phosphate (P) uniquement ou un engrais contenant de l azote et du phosphore (NP). On a détermine la quantite annuelle de P retire e par le grain (toute la paille a e té enfouie) et la variation de la concentration du P disponible dans le sol (couche de 0 a` 15 cm) d apre` s la me thode Olsen au bicarbonate. En 1993, les auteurs ont aménagé des parcelles secondaires qui n ont plus rec u d engrais P, afin de déterminer l effet re siduel des engrais P appliqués au cours des 27 anne es ante rieures. Aux endroits ou` de l engrais P a été appliqué chaque anne e, le grain assimilait en moyenne 54 a` 78 % du P des engrais, avec des pics pouvant atteindre jusqu a` 65 et 109% de 1994 a` 2005, pe riode caracte rise e par la carence en eau la plus faible. Le grain qui poussait sur les parcelles ne recevant que de l engrais P a retire 13 % moins de P du sol que celui des parcelles bonifie es avec l engrais NP. Dans les syste` mes amende s uniquement avec de l engrais P, la concentration de P mesure e selon la technique Olsen augmente line airement dans le temps, mais elle atteint un maximum au bout de 20 à 22 ans avant de se stabiliser, dans les syste` mes recevant de l engrais NP. Le bilan cumulatif de P (P de l engrais moins P absorbe par le grain) explique 60 % de la variabilite de l accumulation de P établie par la me thode Olsen durant l expe rience. Avec la MB, la teneur en P du sol mesure e avec la Corresponding author ( zentnerr@agr.gc.ca). Abbreviations: CW, continuous wheat; FWW, fallow-wheatwheat; NP, nitrogen plus P; PUE, P use efficiency; WD, water deficit Can. J. Soil Sci. (2011) 91:3952 doi: /cjss

2 40 CANADIAN JOURNAL OF SOIL SCIENCE technique Olsen augmente de 0,15 kg par hectare pour chaque kg par hectare de P ajouté en sus de la quantite assimile e par la culture. Le taux d accumulation du P mesure par la me thode Olsen e tait plus éleve (0,20 kg par hectare pour chaque kg par hectare d engrais P en exce dent) chez l assolement JBB, peut-eˆ tre a` cause de la mine ralisation du P durant l anne e de jache` re estivale. Quand on a suspendu l application d engrais P de 1994 à 2005, la production totale de grain a le ge` rement diminue (de 10 %) pour la MC, sans que l assolement JBB s en ressente. La quantite de P absorbe e par la culture ( ) qui n avait pas rec u d engrais P au cours des 12 dernie` res années e quivalait respectivement a` 105 et à 90 % de l engrais P applique aux syste` mes NP et P. Les auteurs en de duisent que les re sidus de P pre sents dans le sol des Prairies sont assimilables par les plantes. Le ble pourrait donc récupérer pre` s de la totalite de l engrais P appliqué au sol si on lui en laisse le temps. Mots clés: Production agricole, bilan P, me thode Olsen de dosage du P, efficacite de l utilisation des engrais P, tchernozem brun After N, P is the nutrient most likely to limit crop growth in the semiarid region of the Canadian prairies. While some producers base their fertilizer decisions on soil tests of individual fields, the vast majority rely on general or regional nutrient recommendations, especially for P. In the dry Brown soil zone, where summer fallow and spring wheat dominate the cropping mix (Campbell et al. 2002), rates of N application to crops grown on summer fallow are low, because significant N mineralization occurs during this 21-mo period and yield potential is generally low due to moisture limitations; however, P fertilizer is usually required. For crops grown on stubble, N application is generally required for optimum yield, while P fertilizer may also be applied. Because of uncertainty about yield responses to fertilization resulting from water limitations, producers often reduce their fertilizer inputs as a way to reduce cost and financial risk. When difficult economic times arise, producers choose between reducing pesticide or fertilizer inputs, and fertilizer (especially P) is often the first input sacrificed. Most prairie soils are high in total P, but low in available P (Doyle and Cowell 1993). Low soil water and low early-spring temperatures further limit P acquisition by crops (Stewart and Karamanos 1986). In Saskatchewan, soils containing less than 10 to 15 kg P ha 1 as Olsen P may give economic yield responses to P each year (Stewart and Karamanos 1986). However, even soils testing high in Olsen P (e.g., 45 kg ha 1 ) can give yield responses to seed-placed P when soil moisture and temperature are low in the month after seeding ( popup or starter-p effect) (Stewart and Karamanos 1986; Roberts 1992). In Saskatchewan (and Manitoba) the recommended rates of P for soils with Olsen P (0- to 15-cm layer) of B10, 1115, 1620, 2160, and 60 kg ha 1 are 13, 11, 9, 7 and 0 kg P ha 1, respectively (Saskatchewan Agriculture 1985). Phosphorus use efficiency is often improved by increasing the supply of other nutrients (Sheppard and Racz 1980; Harapiak and Penney 1984; Grant et al. 1984). For example, maximum crop response to residual P is obtained only when N supply is adequate (Roberts and Stewart 1987). Based on 30 yr of research at the University of Saskatchewan, Halstead (1971) concluded that P concentration in grain was reduced as N rate was increased (likely due to the dilution effect); however, P concentration in grain was unaffected by P rates. In prairie soils, 75 to 90% of fertilizer P is not used by the first crop after application (Spinks and Barber 1946), but some of the residual P is recovered by subsequent crops (Spratt and Read 1980; Roberts 1992). The regular addition of small amounts of P fertilizer over extended periods can result in a large build-up of residual P (Spratt and McCurdy 1966; McKenzie and Roberts 1990; Campbell et al. 2005). On a thin Black Chernozem at Indian Head, Saskatchewan, Spratt and McCurdy (1966) showed that P fertilizer applied once every 3 yr to a summer fallow-wheat-wheat (Triticum aestivum L.) (FWW) system resulted in ever-increasing yields and Olsen P over time, while yields and Olsen P decreased with time when P was not applied. Similar results were obtained in two 36-yr studies, one by Ridley and Hedlin (1962) on a Red River clay in Manitoba, and another by Campbell et al. (2005) on a medium-textured Brown Chernozem in southwestern Saskatchewan. There is strong evidence that residual P in the prairie soils continues to be available to crops over time. In a greenhouse experiment, Read et al. (1973) determined that 19 crops of barley (Hordeum vulgare L.) and oats (Avena sativa L.) grown in soils sampled 3 yr after large applications of superphosphate (100 to 400 kg P ha 1 ) recovered 70 to 87% of the applied P; however, recovery in the field was much smaller because yields were limited by water availability (Spratt et al. 1980). Sadler and Stewart (1974), in an exhaustive literature review, concluded that a substantial proportion of residual fertilizer P remains in forms available to succeeding crops; these conclusions were corroborated by Doyle and Cowell (1993) based on evidence from field and laboratory research. Work conducted at Swift Current, Saskatchewan on a silt loam showed that, in a FWW rotation fertilized with N and P for 24 yr, 60% of the accumulated fertilizer P was in plant-available forms (sorbed P and, to a lesser extent, microbial P); in a continuous-wheat rotation that was also fertilized with N and P, only 31% of the applied P accumulated in the soil (Selles et al. 1995). However, as indicated by the authors, because of the higher organic C returns to CW(NP), it is possible that in this Brown Chernozemic soil a substantial amount of

3 SELLES ET AL. * LONG-TERM EFFECTS OF P FERTILIZATION 41 P may be moved below the 15-cm sampling depth in association with fulvic acids (Schoenau and Bettany 1987). Another study at Swift Current demonstrated that when 17 or 27 kg P ha 1 was applied in a concentrated band, the concentration of P in soil solution was controlled by sorption processes while, at application rates greater than 50 kg P ha 1, Caphosphates controlled the solution concentration of P (Selles 1993). As the cost of fertilizer increases with increasing demand and costs of production and transportation, it is important for producers to ensure that the P applied is used efficiently by crops, and that it is not lost from their farms, or, when left in the soil, it is not converted to forms unavailable to crops. Losses of P from agricultural soils have been identified as a major cause of eutrophication of surface waters (Sharpley and Rekolainen 1997; Leone et al. 2008) and limits to P application may be imposed where land is close to water courses (e.g., Lake Winnipeg Stewardship Board 2005). A long-term crop rotation-fertility field experiment at Swift Current, Saskatchewan, in which wheat yields and Olsen P were monitored annually ( ), provided the basis for this study. The objectives were:(i) to determine the effect of cropping frequency and N fertilization on long-term P use efficiency by wheat, and (ii) to evaluate the effectiveness of residual P (P accumulated over 27 yr of fertilization) as a source of P for wheat. MATERIALS AND METHODS Experimental Design and Crop Management The experiment was established in 1967 on slightly sloping land (B3%) at the Semiarid Prairie Agricultural Research Centre (SPARC) of Agriculture and Agri- Food Canada, near Swift Current (lat ?N, long ?W), Saskatchewan. The site had been cropped since 1922 to a FW rotation with little addition of fertilizer. The soil is a Swinton loam (Ayres et al. 1985), an Orthic Brown Chernozem (Canada Soil survey Committee, Subcommittee on Soil Classification 1978). The ph (water paste) of the top 15 cm was 6.5. Twelve crop rotation-fertility treatments were established on 81, 0.04-ha plots arranged in a randomized complete block design with three replicates. All phases of each rotation were present every year and each rotation was cycled on its assigned plots. In this paper, we discuss four of the original treatments, i.e., FWW and CW rotations, each fertilized either with N and P, or with P alone. We designate these treatments as FWW(NP), FWW(P), CW(NP), and CW(P). In 1993 all plots of these treatments were split to provide an area of 15 by 35 m, where P fertilization was discontinued, while P fertilization was continued in the rest of the plot area. Thus, starting in 1993 we had eight rotationfertilizer regimes. To facilitate field operations, the areas where P fertilization was halted were all located in the southwest corner of the plots. Original Plot Management. Seedbed preparation, herbicide application, seeding, harvesting, and tillage operations were described previously (Campbell et al. 1983, 1992; Zentner and Campbell 1988). Wheat was generally seeded in mid- May at the recommended rate of 67 kg ha 1 using a hoe-press drill with 17.8-cm row spacing. Recommended cultivars were planted each year, but cultivars were changed as new ones became available (Campbell et al. 2005). Commercial farm equipment was used to perform all cultural and tillage operations. Weed control was achieved by a combination of mechanical tillage (mainly cultivator and rodweeder) and herbicides (as required) using recommended methods and rates (University of Saskatchewan 1975). In the fall, after harvest, 2,4-D was applied to all plots to control winter annual weeds. On average, summer fallow plots received approximately four tillage operations. In-crop weed control generally involved the use of bromoxynil plus MCPA E (1:1) and tralkoxydim. Fertilizer N, as (ammonium nitrate), was broadcast before seeding based on amounts of soil NO 3 -N (0- to 60-cm layer) measured in individual plots in the previous fall. From 1967 to 1989 we used N rates recommended by the soil-testing laboratory of the University of Saskatchewan (Saskatchewan Agriculture 1985) with N applied to bring total mineral N (soil testfertilizer N) to 65 kg ha 1. Starting in 1990, rates of N fertilization were changed to follow the new Saskatchewan Soil Testing Laboratory recommendations for N fertilization (N recommendations were increased to 90 kg ha 1 of total N for wheat grown on summer fallow and to 73 kg ha 1 for wheat grown on stubble). Phosphorus fertilizer (monoammonium phosphate) and seed were dispersed through the same tube of the drill (seed-placed). On average, all cropped treatments designated to receive P received 910 kg P ha 1 yr 1. Wheat grown on summer fallow received approximately 8 kg N ha 1 yr 1 up to 1991 and, thereafter, approximately 41 kg N ha 1 yr 1 [partly due to more favourable weather conditions and partly due to the new fertilizer recommendation guidelines (Campbell et al. 2005)]. Wheat grown on wheat stubble received approximately 1032 kg N ha 1 yr 1 in the first 24 yr and approximately 50 kg N ha 1 yr 1 in the final 15 yr. In 1980 and 1982, N was inadvertently applied to the CW(P) treatment at rates of 70 and 40 kg N ha 1, respectively. Wheat was harvested at the full-ripe stage. Yield was determined by cutting a swath 5 m wide and 40 m long through the middle of all four original cropped treatments and the grain was harvested with a conventional combine. Small areas (2.32 m 2 ) were also hand harvested on each plot to determine N and P concentrations in

4 42 CANADIAN JOURNAL OF SOIL SCIENCE grain and straw. All straw was chopped and spread on the plots. The grain was dried at 708C, ground with a Wiley- Thomas mill (Thomas Scientific, Swedesboro, NJ) to B1 mm and analyzed for P using the indophenol procedure from 1968 to 1986 (Varley 1966). From 1987 onwards, plant P was measured by the Automated Acid Molybdate/ANSA procedure (Milbury et al. 1970) using a Technicon analyzer. Soil samples (0- to 15-cm layer) were taken from the original plots with a Giddings soil corer each year after spring thaw but prior to commencement of field operations in early spring, and also in mid-october. Two cores (5 cm diameter) were taken per plot and these cores were combined, sieved (2 mm), sub-sampled for determination of gravimetric water content, and the remaining soil was air-dried and analyzed for NO 3 -N and bicarbonate-extractable (Olsen) P concentrations (Hamm et al. 1970). Bulk density measured at the start of the study (1.22 Mg m 3 for the 0- to 15-cm layer) (Campbell et al. 1983) was used to convert NO 3 -N and Olsen P concentrations to area-based units (kg ha 1 ). In the original plots and the subplots from 1994 onward, soil samples for Olsen P determination were taken with the Giddings soil corer from the 0- to 7.5- and 7.5- to 15-cm depths. We did not measure bulk densities but assumed earlier measured values of 1.15 and 1.29 Mg m 3 for these layers, respectively (Biederbeck et al. 1994). P Withholding Area Management In 1993 we selected an area 15 m35 m in size located in the southwest corner of each original plot. In these areas P fertilization was halted. The drill was equipped with a clutch on the fertilizer drive to allow the operator to stop delivery of fertilizer when going over this area so that only wheat seed was delivered. Plant and soil sampling were taken from an area 50 cm long by 71 cm wide (four plant row widths) in the centre of the sub-plot where P was withheld. Plant samples were removed at the full ripe stage by cutting with a sickle at the same height as the cutter bar of the combine used to harvest the rest of the plots (ca. 15 cm). The plants were threshed in a stationary thresher. In the fall (typically mid-september to October), two soil cores (15-mm diameter) were removed with a hand sampling device from the 0- to 7.5-cm and 7.5- to 15-cm depths for determination of Olsen P. An identical sampling area was established in the portion of the plots where P fertilization was continued without modification. This sampling area straddled the same crop rows chosen for the P withholding area. The same sampling areas were used throughout the study. After 1993, grain yields and P concentration in the grain were determined using these samples (the grain was analyzed as described previously). Agronomic management of the P withholding area was in all other respects identical to that of the original plot area. Calculation of P use efficiency Phosphorus use efficiency was calculated using the approach originally proposed by Karlovsky (1981, 1982) for grazed pastures. He suggested that, if soil available P is constant (at steady state), P utilization efficiency can be calculated as the amount of P taken up by pasture plants, expressed as a percentage of total P applied, which includes P added in fertilizer, and P returned in animal excreta and un-utilized herbage. Extending this approach to wheat,% P utilization (U) can be calculated as the amount of P in grain (G) and straw (S c ) as a percentage of the P input, i.e., P added in fertilizer (F), and P returned in straw from the previous crop (S p ) (all in units of kg P ha 1 ): U(GS c )=(FS p )100 (1) In wheat crops, most of the P is partitioned to grain, with straw P usually making up only approximately 12% of total crop P (grain plus straw) (Selles et al. 1995). As S c and S p are both relatively small and of similar magnitude, Eq. 1 can be re-written to give a good approximation of fertilizer P use efficiency (PUE) by wheat: PUE G=F100 (2) Equation 2 corresponds to the balance method of calculating PUE, which was recommended by Syers et al. (2008). For this study, use of Eq. 2 would result in an underestimation of fertilizer P use efficiency because there was generally a trend for soil available P to increase over time (discussed later). In this work we mostly used cumulative PUE, which was calculated as the sum of P removal in the grain divided by the sum of P applied to the crop. This allowed us to compare the PUE of the original plots with those where P fertilization was halted after Statistical Analysis To simplify interpretation of the results, the crop variables were converted to a per-rotation basis. For example, grain yield data were converted to grain production by adding the yield of the cropped phases and dividing by the rotation length; thus, for the FWW rotation, grain production was calculated as the sum of the yields of wheat seeded on summer fallow and on stubble divided by three. For P removed in the grain, we first calculated the P uptake in each rotation phase as the product of grain yield and grain P concentration. Conversion to a rotation basis followed the same procedure as for grain yields. Phosphorus concentration in grain was calculated as the weighted mean of P concentration in each cropped phase, using grain yield as the weighing factor. Olsen P (0- to 15-cm layer) was averaged across cropped phases of the rotations because

5 SELLES ET AL. * LONG-TERM EFFECTS OF P FERTILIZATION 43 preliminary analysis indicated that there were no significant differences (P 0.05) among rotation phases within each rotation-fertilizer combination. To avoid the confounding effect of highly variable annual weather conditions with systematic changes over time, we divided the experiment into three time segments (1967 to 1979, 1980 to 1993, and 1994 to 2005). These periods were characterized by somewhat different weather patterns (Table 1). The Olsen P data were averaged over the time period, whereas the crop yield and P uptake were converted to rotation totals for the period, as sum of the individual years within the period. Data for the original plots from 1967 to 2005 were analyzed as a randomized complete block design with rotation treatments as main plots and periods (year groups) as subplots with JMP 7.0.2, using the restricted maximum likelihood (REML) method (SAS Institute, Inc. 2007). Data for the plots that were modified in 1993 were analyzed as a randomized complete block design with a nested design with rotations and modified P fertility (P added or P withheld) nested within rotations as main plots and periods as subplots. Because our interest was in identifying periods in which changes had occurred, rather than in making predictions, we considered periods as fixed effects. For the modified plots (1993 to 2005), differences between the P added and P withheld treatments within each rotation in each period were evaluated with single degree of freedom comparisons. Trends in soil Olsen P were analyzed using linear regression procedures available through JMP Because we initially detected a curvilinear trend for two of the original rotations, we used a linear and plateau model, with a spline regression (Freund and Littell 2000) and solved with a non-linear regression procedure of JMP To separate the effect of crop rotation-fertilization combinations on the linear portion of the trends, we used linear regression with indicator variables (Freund et al. 2003, Selles et al. 2006). Trends in Olsen P for the 1993 to 2005 period in the treatments where P fertilization was modified were analyzed using a linear regression procedure with a combination of rotations and modified fertilizer P regime (i.e., P withheld and P added) as an indicator variable. Slopes were compared by LSD, calculated based on the standard Table 1. Water deficit means and coefficients of variability during the study Years Water deficit (mm) CV z (%) 1967 to a to ab to b to z Coefficient of variability for water deficit. a, b Means followed by the same letter are not significantly different (P0.05). errors of the estimate of the trend and the interaction of time with the treatment, assuming that the trend estimate when comparing trends across treatments is given by the algebraic sum of the overall trend of all treatments plus the deviations of the trend of each treatment from the overall trend. The uncertainty in the treatment trends was calculated using accepted procedures for the error of propagation in calculations. RESULTS AND DISCUSSION Growing Season Conditions To characterize growing season weather conditions we used water deficit (WD), calculated as the absolute difference between plant available water [available spring soil water to 120 cm plus growing season precipitation (May 01Aug. 31)] and Class A Pan evaporation for the same period. Precipitation and evaporation were measured at a meteorological station approximately 0.5 km west of the test site. The mean WD and their coefficients of variability (CV) allowed us to divide the study into three periods of approximately equal length (12 to 14 yr). The WD was highest in 1967 to 1979 (first period) (Table 1), due to consistently high temperatures and evaporation, and low precipitation. In 7 of 13 yr WD was larger than the 0.66 quantile; in 4 yr WD was between the 0.33 and 0.66 quantiles, while in only 2 yr was WD smaller than the 0.33 quantile. From 1980 to 1993 (Period 2), WD was not significantly different (P0.10) than in the first period, but it was more variable, with a CV nearly twice that for the first period (Table 1). There were severe droughts during Period 2 (in 1984, 1985 and 1988), when we registered the highest WD values during the study (data not shown). In 6 of 14 yr during the second period, WD was between the 0.33 and 0.67 quantiles; in 4 yr it was above the 0.67 quantile, and in another 4 yr it was below the 0.33 quantile. Finally, from 1994 to 2005 (Period 3), the period in which N rates were increased, WD was significantly lower than in the previous two periods (Table 1). During the last period, WD was smaller than the 0.33 quantile in 7 of 12 yr, it was between the 0.33 and 0.66 quantile in 3 yr, and only in 2 yr was WD above the 0.66 quantile (data not shown). Grain Production Grain production in the original treatments (Table 2) generally reflected the trend in WD (Table 1), i.e., Period 3 Period 2 Period 1. As expected, production for FWW was less than for CW because of the one-third summer fallow. During Period 1 (driest period), withholding N from CW decreased grain production by 2 Mg ha 1, but there was no effect of withholding N from FWW (Table 2). Withholding N from FWW resulted in significantly (P B0.05) lower grain production in Periods 2 and 3 (1.3 and 4.4 Mg ha 1 less, respectively). Similarly, withholding N from CW in Periods 2 and 3 resulted in 4.9 and 9.7 Mg ha 1 less production,

6 44 CANADIAN JOURNAL OF SOIL SCIENCE Table 2. Grain production, grain phosphorus concentration, phosphorus removal in the grain, fertilizer applied, phosphorus balance, Olsen phosphorus, and phosphorus use efficiency (PUE) for the original plots during the periods and totals for the study, based on combine harvest data Grain production z Grain P concentration y Grain P removal x Fertilizer Olsen P added P balance w (0- to 15 cm) PUE based on total uptake v Period Rotation (Mg ha 1 ) (g kg 1 ) (kg P ha 1 ) (%) 1967 to 1979 CW(NP) 18.0c 3.7e 67.1d b 13.1fg 53.1gh CW(P) 16.0de 3.9cd 62.0de ab 14.8f 42.8h FWW(NP) 14.5f 3.4f 48.0f c 13.0g 57.6efg FWW(P) 13.6f 3.4gf 46.6f c 13.2fg 55.2fgh Period Mean to 1993 CW(NP) 20.0b 4.0c 80.4b c 22.4d 60.7def CW(P) 15.1e 4.2b 63.9de a 25.8b 49.5h FWW(NP) 16.9cd 3.8de 63.4de d 19.0c 71.7c FWW(P) 15.6e 3.8de 59.2e d 19.5e 67.0cd Period Mean to 2005 CW(NP) 25.9a 4.2b 107.8a e 24.1c 93.1b CW(P) 16.2de 4.5a 73.5c c 33.4c 65.4cde FWW(NP) 20.4b 3.9cd 80.4b f 20.3e 108.7a FWW(P) 16.0de 4.0c 64.6de e 25.0bc 86.1b Period Mean LSD 0.05 Rotation LSD 0.05 Period 2.3 LSD 0.05 Rot*Per Totals u CW(NP) 63.9a 3.9b 255.4a b 27ab 68.1 CW(P) 47.3c 4.2a 200.1b a 34a 53.8 FWW(NP) 51.8b 3.7c 192.4b d 22b 78.0 FWW(P) 45.2c 3.8c 170.4c c 30a 68.8 LSD 0.05 Rot Study Mean z Grain production for FWW was sum of fallow-wheat plus stubble-wheat divided by 3; for CW it was the same as yield. y P concentration in grain of FWW was calculated as the weighted mean of P concentration in each cropped phase using yield as the weighting factor. x P uptake of FWW was calculated as the sum of P uptake of each cropped phase (i.e., P concentrationyield) divided by 3. w P balance was calculated as the difference between fertilizer P added and grain P removal. v PUEP uptake in grain for rotation divided by P added as fertilizer, times 100. u For Olsen P totals reports the level observed in 2005, for all other variables reports the totals for the entire experiment. ag Means followed by the same letter are not significantly different (P0.05) within a column. respectively. This reflects the importance of adding N fertilizer to stubble crops [less important for summer fallow crops because significant N is mineralized in the summer fallow period (Campbell et al. 2005)]. Further, these results reflect the positive interaction of N and available water on yield response (Henry et al. 1986). Over the duration of the experiment ( ) grain production was 16.6 Mg ha 1 greater for CW(NP) than for CW(P), and 6.6 Mg ha 1 greater for FWW(NP) than for FWW(P). Furthermore, production for CW(NP) was 12.1 Mg ha 1 greater than for FWW(NP). However, there was no difference in production between CW(P) and FWW(P). Grain Phosphorus Concentration Grain P concentration was higher in CW than in FWW (Table 2), likely due to dilution because grain yields for FWW were higher than for CW (Campbell et al. 2005). Grain P concentration increased over time (Period 3Period 2Period 1), due at least in part to increased amounts of available soil P. Withholding N had no effect on grain P concentration of FWW in any period, but it increased P concentration in CW by 0.2 to 0.3 g kg 1, probably reflecting yield dilution in the N fertilized treatments [grain yield of CW(NP) CW(P)] (Campbell et al. 2005). Grain Phosphorus Uptake Grain P uptake was greater for CW(NP) than for CW(P), and greater for FWW(NP) than for FWW(P) (Table 2). These trends primarily reflect the effect of cropping frequency and N fertility on grain production. Not surprisingly, grain P uptake also increased with decreasing WD (Period 3Period 2 Period 1), reflecting the influence of available water on yield and thus P uptake. Clarke et al. (1990) showed that differences in P uptake by wheat were mainly a function of yield rather than changes in grain P concentration. The increase in P uptake by the addition of N fertilizer for CW occurred in all three periods, with the increase being progressively greater as WD decreased. However, for FWW the effect of N on P uptake was significant (P B0.05) only in Period 3.

7 SELLES ET AL. * LONG-TERM EFFECTS OF P FERTILIZATION 45 During , total P removal in grain ranged from 170 kg ha 1 for FWW(P) to 255 kg ha 1 for CW(NP) and, in general, followed the same ranking as observed for grain production. Over this 39-yr period, P removal in the grain represented between 54 and 78% of the P applied as fertilizer and, in Period 3, grain P removal increased to between 65 and 109% of added P. These high recovery rates support the view that, given enough time, crops will recover most of the P applied in fertilizers under prairie conditions (Syers et al. 2008). The amount of P from straw returned to the soil in 39 yr of cropping ranged from 22 to 33 kg ha 1 (0.56 to 0.85 kg P ha 1 yr 1 ). Straw P returns were greater for CW than for FWW (P B0.0001), reflecting the productivity of the rotations, as discussed above for grain. Olsen Phosphorus Olsen P content in the soil in 2005 (Table 2) reflected the amounts of P applied as fertilizer over the study period (CWFWW because of greater P input to CW) and P removal in grain (P-onlyNP treatments because of the lower P uptake with the P-only treatment). This was especially true in Period 3 due to the lower WD facilitating enhanced grain production and thus P uptake (Table 2). During the first period, there were no differences (P 0.05) in Olsen P to 15-cm depth among the four treatments. However, in Period 2, Olsen P in all treatments increased relative to Period 1 to levels close to (or above) those where responses to applied P are unlikely (Read et al. 1977) (Table 2). At this point, the effects of cropping frequency and fertilization on Olsen P started to become more noticeable; CW had higher Olsen P than FWW, and CW(P) had significantly higher Olsen P than CW(NP) (P 50.05). During Period 3, Olsen P increased relative to Period 2 for all treatments. Phosphorus Balance The P balance was calculated as the difference between fertilizer P added and grain P removal (Table 2). The P balance was lower (P50.05) for FWW than for CW. This was mostly because FWW received 33% less fertilizer P than CW while it took up only 20% less P than CW. The P balance of treatments fertilized with N and P was smaller than for the P-only fertilized treatments, where yields (and thus grain P removal) were limited by lack of N. At the beginning of the study, the differences in P balances between CW(NP) and CW(P) and between FWW(NP) and FWW(P) were not significant (P0.05), presumably because soil N reserves were large enough to supply the requirements of the crops. With time, and as weather conditions became more favourable for crop production (Table 1), and available N in the soil decreased (data not shown), grain production in the treatments fertilized with P alone became progressively N limited and therefore their P uptake declined relative to treatments fertilized with N and P. This was more noticeable for the CW(P) than for the FWW(P) rotation, presumably because N mineralization during the summer fallow year partially alleviated crop N deficiency. During the entire study period ( ), the P balance was 54 and 77 kg ha 1 for FWW(NP) and FWW(P), compared with 119 and 172 kg ha 1 for CW(NP) and CW(P), respectively (Table 2). Soil Olsen P increased steadily throughout the 39 yr in response to positive P balances (Table 2). In the treatments fertilized with N and P, however, the rates of increase were larger at the beginning of the study and declined as grain production and P uptake increased in Period 3 (Fig. 1). Phosphorus Use Efficiency Phosphorus use efficiency increased with available water (Period 3 Period 2 Period 1) and was generally greater for NP than for P-only treatments, especially Fig. 1. Trends in Olsen Phosphorus for the original plots, 1967 to [Trend models given by following expressions: CW(NP) if time 522, y time, otherwise y time0.61(time22); FWW(NP) if time 520 y time, otherwise y time0.59 (time20); CW(P) y time; FWW(P) y time].

8 46 CANADIAN JOURNAL OF SOIL SCIENCE at lower WD (Table 2). Further, PUE was generally greater for FWW than for CW. Over the study period, the relative P uptake for FWW(NP) vs CW(NP) was 75.3% and for FWW(P) vs CW(P) it was 85%, whereas the fallow-containing system received only 66% of the P applied to CW. The greater PUE of FWW vs. CW is likely related mainly to improved water availability to the wheat crop grown after the summer fallow year. Between 1967 and 2005, PUE ranged between 53.8% for CW(P) and 78.0% for FWW(NP). These results suggest that, under Canadian prairie conditions, the recovery and use of P fertilizer by crops is not greatly impeded by soil immobilization processes as has often been assumed historically (Syers et al. 2008). Trends in Olsen Phosphorus in the Original Plots Using mean annual values of Olsen P in the rotations, assessed as the mean of the cropped phases for each treatment, we quantified the temporal trend of Olsen P (Fig. 1). In the rotations fertilized with P alone, Olsen P increased linearly with time for the duration of the study (R , PB0.0001). The rate of increase for CW(P) and FWW(P) was 0.68 and 0.45 kg P ha 1 yr 1, respectively (Fig. 1) (slopes were significantly different at P50.05). The higher rate of increase in Olsen P for CW(P) vs. FWW(P) was related to the much greater rate of fertilizer P (373 vs. 248 kg P ha 1, respectively) compared to P removal in grain (200 vs. 170 kg P ha 1, respectively) during the experiment (Table 2). The rotations fertilized with both N and P displayed a linear trend that persisted for 20 to 22 yr, followed by a period with nearly stable levels of Olsen P (Fig. 1). Regression analysis revealed that in the case of CW(NP), there was a linear increase of 0.64 kg P ha 1 yr 1 (P5 0.05) for the period 1967 to 1989 (95 yr), followed by a non-significant trend (P 0.05) thereafter (Fig. 1). For the FWW(NP) treatment there was a linear increase of 0.56 kg P ha 1 yr 1 (P50.05) from 1967 to 1986 (93 yr), followed by a non-significant trend until 2005 (P0.05). The levelling off of Olsen P in CW(NP) and FWW(NP) after 1990 can be explained by a substantial reduction in the P balances of these two treatments in later years when growing season conditions were more favourable for grain production and P removal in the grain (Table 2). This large increase in P removal by the crop, with no change in P fertilization rate, led to a balancing of input and removal of P in this period, thus reducing the rates of accumulation of residual P. Conversely, the rotations fertilized with P alone were not able to capitalize on the improved growing conditions after 1992 because of N deficiencies, thus they continued to have a large excess of soil P relative to crop needs. Consequently, CW(P) and FWW(P) had significantly higher P balances than the corresponding rotations fertilized with N and P (P50.05), and their levels of soil available P continued to increase throughout the experiment (Fig. 1). Olsen P trends among the various rotations between 1967 and 1987 (the period in which all rotations exhibited a linear trend) were evaluated with a linear regression model using crop rotation treatment as an indicator variable to determine the deviation of the trend of individual rotation treatments from the average trend (Table 3). This regression analysis (R , PB0.0001) revealed that, in CW(P), Olsen P increased at a rate of 0.78 kg P ha 1 yr 1, which was significantly larger (P50.05) than the rate in any other rotation (the mean rate of increase for the other rotations was 0.54 kg P ha 1 yr 1 ). The magnitude of the trend in Olsen P was proportional to the total P balance for each rotation from 1967 to The partial correlation matrix in a multivariate correlation between Olsen P and a number of variables likely to influence the Olsen P value revealed that the total P balance between the start of the study and the date of soil sampling for Olsen P (i.e., cumulative rotation treatment P balance) was the variable that accounted for the largest proportion of the variability in Olsen P (data not shown). A simple linear regression between Olsen P and cumulative P balance accounted for 60% of the variability of Olsen P; however, cropping frequency also had an influence on the slope of the regression line (data not shown). When we included cropping frequency as an indicator variable in the regression model, it explained 66% of the variability in Olsen P. This model indicated that in the FWW treatments Olsen P increased by 0.20 kg ha 1 for each kg ha 1 of P added in excess of crop removal (P B0.0001), while in the CW treatments the rate of change in Olsen P was only 0.15 kg ha 1 (PB0.0001) (slopes were significantly different at P 0.006) (data not shown). The rate of Olsen P increase in the FWW treatments was nearly the same as that observed in a study of residual fertilizer P under summer fallow-wheat Table 3. Olsen phosphorus trends by rotation from 1967 to 1987 (period of linear trend for all rotations) determined by a regression model with indicator variables Average trend Rotation shifter z Trend Total P balance Rotation (kg P ha 1 yr 1 ) (kg P ha 1 ) CW(NP) b (NS) 0.60b 97a CW(P) a (0.002) 0.78a 105a FWW(NP) b (.06) 0.48b 55b FWW(P) b (NS) 0.51b 58b LSD z Shifter refers to the deviation of a rotation slope from the overall slope for all treatments. a, b Values followed by the same letter are not significantly different (P0.05) within a column. Values in parentheses indicate probability of the shifter estimate being different from zero; NSnot significant (P0.1).

9 SELLES ET AL. * LONG-TERM EFFECTS OF P FERTILIZATION 47 cropping in a similar soil type in Montana, USA (Halvorson and Black 1985). The higher slope of the FWW regression suggests that, during the summer fallow period, there was a release of soil P to plantavailable P form (Olsen extractable) possibly due to mineralization of organic P. Effect of Withholding Phosphorus Fertilization with P for 27 yr (1967 to 1993) in amounts larger than P removal in the grain led to a significant increase (P50.01) in Olsen P in both rotations from suboptimal levels to levels considered adequate for wheat (Read et al. 1977). Consequently, although we observed a few significant effects (P B0.05) of withholding P on grain production and P uptake in the CW treatments, these effects were generally sporadic and more related to variations in weather conditions, than to a consistent trend induced by inadequate P supply. Between 1994 and 2005 total grain production decreased by 10% in the CW systems with no P fertilization, but there was no effect for FWW (Table 4). However, the yield reduction in CW was not consistent, but was the result of 40 and 34% reductions where P was withheld from CW(NP) in 2001 and 2002, respectively, and reductions of 36 and 43% in 2000 and 2002, respectively, under CW(P). In no other year was there a significant (P 50.05) reduction in grain production where P was withheld (data not shown). Water deficit in 2001 was the fourth highest during the study, and it was significantly higher (P ) for the CW (970 mm) than for FWW (933 mm). We speculate that the growth of wheat in the original CW treatments was favoured by the seed-placed application of P, allowing the crop to initiate rapid root growth that facilitated a more thorough exploration of the soil profile at a critical stage of development. In contrast, where P was withheld, the crop probably had a smaller root system and was not able to exploit soil water reserves as fully as the crop that received P. Under FWW, withholding P did not decrease production, perhaps because extra P was released during the summer fallow period, and this allowed the crop to achieve good early root development similar to that induced by seed-placed fertilizer P. Phosphorus removal in the grain was determined mainly by grain yield as found by others (Clarke et al. 1990). Regardless of the short-term P management, the rotations fertilized with N removed significantly more P (P50.05) than those receiving no N and, as for grain production, P uptake was larger (P50.05) in CW than in the FWW rotations (Table 4). As observed with grain production, withholding P did not affect grain P uptake in the FWW treatments, but it significantly decreased uptake of P in CW(NP) and CW(P) (P and P0.03, respectively) by 16 and 10 kg P ha 1, respectively. This difference was only in the period (data not shown). Phosphorus concentration in the grain was not affected by short-term P management (P 0.21) (data not shown). Phosphorus Balance Between 61 and 132 kg P ha 1 accumulated in the soil between 1967 and 1993 (Table 4), but there was little or no soil P accumulation (10 and 21 kg ha 1, respectively) in P-treated plots between 1994 and Improved water availability in this period (Table 1) resulted in an average 62% increase in the annual grain P uptake (Table 2). The negative P balances (106 to 68 kg ha 1 ) for the treatments receiving no P for the last 12 yr indicated that in this period the crop used residual P derived from fertilizer applied in the previous 27 yr. Our results support previous studies indicating that in the semiarid prairies residual fertilizer P remains in a form that is available to subsequent crops (Sadler and Stewart 1974; Spratt et al. 1980; Doyle and Cowell 1993). Our crop P recovery data are also consistent with Selles et al. (1995), who showed that most of the residual P was in chemically-labile forms. At the end of the study in 2005, P balances continued to be positive in the treatments receiving P for the entire 39 yr; in these plots P fertilization exceeded P removed in grain by 51 to 154 kg ha 1. However, where P fertilization was halted for the last 12 yr, the long-term P balances were much smaller, ranging from 20 kg ha 1 under FWW(NP) to 51 kg ha 1 under CW(P). Under CW(NP) and FWW(P) the final balance in 2005 was not different from zero (P 0.05), indicating that in these two treatments the equivalent of all the residual P accumulated in the soil in the first 27 yr was removed by the crop in the 12 yr when P was withheld. Under FWW(NP) the P balance was negative (P 50.05), i.e., the crop was able to remove the equivalent of all of the fertilizer-derived P plus another 20 kg ha 1 of soil P. Finally, under CW(P), the crop under prolonged N stress was able to remove only approximately 100 kg ha 1 of P, leaving unused the equivalent of approximately 50 kg ha 1 of fertilizer P. Changes in Olsen Phosphorus when Phosphorus was Withheld Levels of Olsen P were affected by rotation-fertility treatment and by cessation of P fertilization after 1993 (Fig. 2). The trend in Olsen P after 1993 was assessed with a linear regression model as a function of time and P management (original treatments vs. P withheld) as indicator variables. This analysis showed that these two factors accounted for 77% of the variability in Olsen P (PB0.0001) (Table 5 and Fig. 2). Consistent with the trend analysis for the original treatments discussed earlier (Fig. 1), the two rotations fertilized with P alone exhibited positive trends for Olsen P (P50.05) when P fertilization was continued, and small, though significant (P 50.05), negative trends when P was withheld (Table 5). The rotations fertilized with both N and P showed no trend in Olsen P when we continued to apply P [as previously discussed (Fig. 1)]. When P was withheld, Olsen P in CW(NP) and FWW(NP) showed strong negative trends (P B0.0001) that were much steeper (P 50.05) than the trends of the corresponding

10 Table 4. Effect of withholding phosphorus on grain production, phosphorus removal in the grain, phosphorus balances, and phosphorus use efficiency (PUE) Grain production Grain P removed P fertilizer applied P balance PUE P applied P withheld P applied P withheld P applied P withheld P applied P withheld P applied P withheld Period Rotation (Mg ha 1 ) (kg ha 1 ) (%) z CW(NP) CW(P) FWW(NP) FWW(P) LSD 0.05 Rot CW(NP) x ** CW(P) x ** FWW(NP) FWW(P) LSD 0.05 Rot 15 LSD 0.05 MF[Rot] y CW(NP) x x CW(P) ** FWW(NP) FWW(P) LSD 0.05 MF[Rot] 2 19 NA 9 6 z Data for grain production, P uptake, and P balances in based on combine yields; in based on hand sampling. y Microplot fertility nested within rotation. x Single asterisk indicates significant differences between P applied and P withheld plots as determined by single degree of freedom contrasts at P50.05, double asterisk at P CANADIAN JOURNAL OF SOIL SCIENCE