Do Rates and Splitting of Phosphogypsum Applications Influence the Soil and Annual Crops in a No-Tillage System?

Applications of phosphogypsum (PG) provide nutrients to the soil and reduce Al activity, favoring soil fertility and root growth, but allow Mg mobilization through the soil profile, resulting in variations in the PG rate required to achieve the optimum crop yield. This study evaluated the effect of application rates and splitting of PG on soil fertility of a Typic Hapludox, as well as the influence on annual crops under no-tillage. Using a (4 × 3) + 1 factorial structure, the treatments consisted of four PG rates (3, 6, 9, and 12 Mg ha) and three split applications (P1 = 100 % in 2009; P2 = 50+50 % in 2009 and 2010; P3 = 33+33+33 % in 2009, 2010 and 2011), plus a control without PG. The soil was sampled six months after the last PG application, in stratified layers to a depth of 0.8 m. Corn, wheat and soybean were sown between November 2011 and December 2012, and leaf samples were collected for analysis when at least 50 % of the plants showed reproductive structures. The application of PG increased Ca concentrations in all sampled soil layers and the soil pH between 0.2 and 0.8 m, and reduced the concentrations of Al in all layers and of Mg to a depth of 0.6 m, without any effect of splitting the applications. The soil Ca/Mg ratio increased linearly to a depth of 0.6 m with the rates and were found to be higher in the 0.0-0.1 m layer of the P2 and P3 treatments than without splitting (P1). Sulfur concentrations increased linearly by application rates to a depth of 0.8 m, decreasing in the order P3>P2>P1 to a depth of 0.4 m and were higher in the treatments P3 and P2 than P1 between 0.4-0.6 m, whereas no differences were observed in the 0.6-0.8 m layer. No effect was recorded for K, P and potential acidity (H+Al). The leaf Ca and S concentration increased, while Mg decreased for all crops treated with PG, and there was no effect of splitting the application. The yield response of corn to PG rates was quadratic, with the maximum technical efficiency achieved at 6.38 Mg ha of PG, while wheat yield increased linearly in a growing season with a drought period. Soybean yield was not affected by the PG rate, and splitting had no effect on the Division – Soil Use and Management | Commission – Soil Fertility and Plant Nutrition Vicensi et al. Do Rates and Splitting of Phosphogypsum Applications Influence the Soil... 2 Rev Bras Cienc Solo 2016; v40:e0150155 yield of any of the crops. Phosphogypsum improved soil fertility in the profile, however, Mg migrated downwards, regardless of application splitting. Splitting the PG application induced a higher Ca/Mg ratio in the 0.0-0.1 m layer and less S leaching, but did not affect the crop yield. The application rates had no effect on soybean yield, but were beneficial for corn and, especially, for wheat, which was affected by a drought period during growth.


INTRODUCTION
Brazilian agriculture has an outstanding grain production, achieved by technologies that maximize the yield potential, such as the implementation of no-tillage (NT) systems. Since the introduction in the 1970s (Derpsch and Friedrich, 2009), NT proved to be more than a set of practices that ensure minimum soil disturbance, soil mulching with crop residues and control of soil erosion. The system improves the soil quality as a whole, favoring yield and allowing greater resource use efficiency (e.g. of soil, fertilizers and fuel), which has promoted the expansion of NT across the country. Currently, in Brazil, an acreage of nearly 32 million hectares is being managed in NT systems (FAO, 2012).
To establish well-nourished crops, the soil must retain water and nutrients in readily available forms for uptake, and no limiting factors of soil chemistry, physics and biology may stress the crop (Klein, 2011). In agriculture, short-term stresses during critical stages of the plant life cycle can greatly affect the yield of staple crops, especially due to the high precocity of current varieties. In spite of knowing NT since the 1970s, many farmers now restrict this practice to "direct seeding on crop residues" rather than applying the complete "no-tillage management system" with principles of conservation agriculture. This has affected the soil, resulting in stratified properties; satisfactory chemical conditions tend to be restricted to the topsoil, while compaction increases in the subsurface layers, intensifying the cropping problems resulting from the poor weather conditions in recent years (Denardin et al., 2008;Nunes et al., 2014).
Since in Brazil and other South American countries NT is used uninterruptedly, there is no periodic soil tillage, the accumulation of crop residues in the topsoil and the addition of lime and fertilizers to the surface intensifies soil stratification over time, leading to the accumulation of organic C and N (Ebeling et al., 2008;Venzke Filho et al., 2008), as well as P, K + , Ca 2+ and Mg 2+ to depths of 0.0-0.10 m (Klein et al., 2007;Spera et al., 2011;Acqua et al., 2013;Kramer et al., 2014). Since the vertical mobility of lime is reduced in NT, the subsurface soil changes to a more acidic environment with higher concentrations of exchangeable aluminum (Al 3+ ) (Raij et al., 1998;Ernani et al., 2001), creating a chemical limitation to root growth.
Management strategies originally designed from the fundamentals of NT can minimize or mitigate the problems resulting from the vertical gradient of fertility in the soil; implementing crop diversification, reducing time between sowing and harvesting and the addition of crop residues in a manner compatible with the biological dynamics (Nunes et al., 2014) can all produce good results in the medium or long-term. Phosphogypsum (PG), on the other hand, even if applied to the soil surface, improves the chemical conditions in the soil profile for root growth, especially in deeper layers (Carvalho and Raij, 1997), after a few months (Caires et al., 2001, Rampim et al., 2011. Due to its solubility and other specific reaction characteristics, PG (CaSO 4 .2H 2 O) moves throughout the soil profile releasing Ca 2+ and sulfur (S) (Caires et al., 2006(Caires et al., , 2011aSoratto and Crusciol, 2008). Concomitantly, Mg 2+ and K + -the latter only rarely -are displaced from the soil exchange sites in top layers and then move along with water to the subsurface (Caires et al., 2011b). In addition, a decrease in the Al 3+ concentrations to levels not toxic Rev Bras Cienc Solo 2016;v40:e0150155 to plant roots can occur (Toma et al., 1999), due to the formation of ionic pairs between Al 3+ and the sulfate (SO 4 2− ) or fluoride (F − ) contained in PG (Zambrosi et al., 2007).
Although comprehensive information regarding the effects of gypsum on soil and plants is available, Brazil has not yet established criteria for recommending PG application rates, which can vary regionally and according to the local climate, soil order and crop. There are few available data regarding crop rotation systems, and no information about PG split-application rates, which is a common strategy for increasing the efficiency of fertilizers by reducing N and K fertilization losses (Cardoso et al., 2007).
The hypotheses of this study were that the higher availability of Ca 2+ and S and the decrease in Al 3+ in the soil due to PG application improve annual crop yields under long-term NT, and that PG split applications reduce the loss and, or, mobilization of S, Mg 2+ and, or, K + , especially at higher PG rates. The objective was to evaluate the effects of different PG rates and split applications on a Typic Hapludox chemical properties, nutrition and on yields of NT corn, wheat and soybean.

MATERIALS AND METHODS
This study was carried out between November 2011 and April 2013, as part of a long-term field trial established in 2009 at the Experimental Field of the West-Central State University, in Guarapuava, Paraná, where the climate (Köppen-Geiger System) is classified as Cfb, mesothermal humid subtropical. A meteorological station of the Agronomic Institute of Paraná (IAPAR), located 200 m away from the experimental site (25° 23' S, 51° 30' W and 1,026 m above sea level), was used to collect rainfall and temperature data throughout the experimental period, as well as long-term averages ( Figure 1). In October 2009, a trench was dug for the morphological assessment of the soil and for soil sampling to determine the clay content (Claessen, 1997) and soil chemical properties (Pavan et al., 1992).
The profile (Table 1) was classified as very clayey Latossolo Bruno Distrófico, based on the Brazilian soil classification system (Santos et al, 2013), a Typic Hapludox (Soil Survey Staff, 2014. More information about the initial phase of the experiment was provided by Michalovicz (2012).
A randomized complete block design was used, with four replications and 16 × 6.4 m plots. The PG rates constituted the first treatment factor: 3, 6, 9, and 12 Mg ha -1 (dry-weight basis). The second treatment factor consisted of the split applications: P1 -without splitting, i.e., 100 % applied in November 2009; P2 -application split in two years, 50+50 % in November 2009 and 2010; and P3 -split amongst three years, 33+33+33 % in November 2009, 2010 and 2011. A control without PG was also part of the study, in a factorial arrangement (4 × 3) + 1. The Ca and S concentrations of PG were 170 and 140 g kg -1 , respectively, and the application rates were calculated to provide 0, 33, 66, 100, and 133 % of the required amount of Ca 2+ in the A1 horizon (Table 1), so as to reach 60 % of Ca 2+ saturation of the cation exchange capacity (CEC pH 7.0 ).
The factorial structure was complete when the final application of PG, the third 33 % portion in P3, was applied to the soil surface on November 15 th , 2011, together with corn (P3646H®) sowing. The final plant density was 62,500 plants ha -1 , in a row spacing of 0.8 m, and the crop was fertilized with 300 kg ha -1 of 14-33-00 N-P-K fertilizer mixture, applied in the planting furrow. Side-dressing consisted of 45 kg ha -1 N (urea) and 54 kg ha -1 K 2 O (KCl) at V4, plus 58 kg ha -1 N (urea) at V6. Wheat (OR Mirante®) was sown on July   Soil samples were collected in May 2012, after the corn harvest (six months after the last PG application). A composite sample, consisting of 12 subsamples per plot (four from the planting rows and eight from in-between the rows), was collected with an auger probe from the layers 0.0-0.1 and 0.1-0.2 m. At six of these points (two in and four in-between the rows), the soil was also sampled at depths of 0.2-0.4, 0.4-0.6 and 0.6-0.8 m with a Dutch auger. Samples were dried in a forced-air oven at 65 °C, and then ground in a knife-type mill and sieved through a 2 mm mesh. The analysis was done according to Pavan et al. (1992) and P extracted by Mehlich-1. The inorganic form of S (SO 4 2− ) was extracted by calcium phosphate 0.01 mol L -1 as described by Cantarella and Prochnow (2001), and then determined through turbidimetry. The soil delta pH (∆pH) was also calculated, as described by Kiehl (1979).
The yield of the crops harvested manually at physiological maturation was evaluated by collecting four sub-samples of 4 m (16 linear meters) per plot for corn and 3 m (12 linear meters) per plot for wheat and soybeans. Yields were reported at a moisture of 130 g kg -1 and maximum technical efficiency (MTE) rates of PG were obtained by setting the first derivative of the yield response models of each crop (regression analysis) equal to zero. All results were subjected to analysis of variance and in the case of significant (p<0.05) interactions between factors regression analyses were conducted separately (partitioning) for each split treatment. When no interaction was observed, split treatment means were tested by Tukey's test (α=0.05) and regression analyses were conducted using the means of each rate for all split treatments. The models with the highest level of significance were adopted.

Chemical properties of soil fertility
The PG rates did not influence soil pH in the 0.0-0.1 and 0.1-0.2 m layers, with or without split applications (Table 2). However, soil pH increased in the 0.2-0.4 and 0.4-0.6 m layers when compared with the control treatment, regardless of total or split application rates, while a linear increase in soil pH was observed in the 0.6-0.8 m layer as a function of PG rate, without effect of splitting or interaction. Phosphogypsum is a neutral salt without corrective properties for soil acidity, so no change in soil pH was expected from its use. Notwithstanding, there can be a small magnitude pH increase in subsurface soil due to SO 4 2− -S increase, which at high concentrations can displace the OH − that is adsorbed to Fe and Al hydrated oxides to the soil solution, a process known as "self-liming" (Reeve and Sumner, 1972), which is rather unstable and reversible (Raij, 2008). Similar results were obtained by Caires et al. (1999) for the 0.2-0.4, 0.4-0.6 and 0.6-0.8 m layers, and in another study for the 0.2-0.4 and 0.4-0.6 m layers (Caires et al., 2003).
Another possible explanation for these increases in soil pH could be a higher absorption of nitrate (NO − 3 ) in deeper layers, since NO − 3 is highly mobile in the soil profile and one of the plant effects described for PG is a greater root distribution in deeper soil layers (Caires et al., 2001). The OH − is released by plants when NO − 3 is absorbed, alkalizing the soil (Bloom et al., 2003).
The application of PG resulted in a decrease in the Al 3+ concentration in all studied soil layers, however, there was no effect of application splitting or rate increase ( Table 2). The effects of PG on Al 3+ have been associated with the formation of Al hydroxylated structures in the soil, as a result of ligand exchange between OH − from the surfaces of hydrated Fe or Al oxides and SO 4 2− provided by PG (Reeve and Sumner, 1972), as well as the formation of an ionic pair between Al 3+ and SO 4 2− or fluoride (F − ) (Zambrosi et al., 2007). Phosphogypsum contains approximately 15 % S and 0.6-3.2 % F, and even when considering the role of SO 4 2− in Al 3+ complexation, F − may be even more effective in this reaction (Carvalho and Raij, 1997;Zambrosi et al., 2007), which may explain why PG is more efficient than natural gypsum (gypsite) in complexing Al 3+ . In clayey Oxisols, Rampim et al. (2011) and Caires et al. (1999) also reported a reduction in Al 3+ concentrations due to PG application, to a depth of 0.8 m in the latter case.  (2) Linear adjust and R 2 value after the bar; * and **: significant at p<0.05 and p<0.01, respectively; ns: not significant.
Neither PG application rates nor splitting had an effect on H+Al concentrations in any of the soil layers (Table 2), in agreement with the observations of Caires et al. (2004) under similar soil conditions. Although there were changes in Al 3+ and pH concentrations in this study, these were minor and not intense enough to affect the H+Al values, which is consistent with PG being a neutral salt.
The Ca 2+ concentrations increased linearly in response to the PG application rates in all soil layers, but no splitting effect was found (Table 3). Phosphogypsum consists of 17 % Ca, and given that the Ca 2+ concentrations were found to have increased in all soil layers, this demonstrates that PG is a Ca 2+ source that migrates downwards through the soil profile. An increase in Ca 2+ concentration throughout the soil profile was also observed by Caires et al. (1999Caires et al. ( , 2003Caires et al. ( , 2011b, Soratto and Crusciol (2008) and Rampim et al. (2011) during their studies, noting that PG is a complementary soil amendment tool to lime for increasing Ca 2+ concentrations in NT soils, especially in the subsurface, due to the higher PG solubility and vertical mobility in the soil profile compared with lime (Caires et al., 1998).
Magnesium concentrations decreased linearly with increasing PG application rates in the 0.0-0.1 and 0.1-0.2 m layers, as reported in other studies (Toma et al., 1999;Soratto and Crusciol, 2008;Caires et al., 2011a), and again no effect of splitting was observed. Competing for the same adsorption sites, the Ca 2+ added by PG displaces soil Mg 2+ ions to the soil solution, which is further transported to deeper soil layers as the water percolates through the profile (Medeiros et al., 2008). The formation of MgSO 0 4, which is mobilized easily in the soil, also occurs (Zambrosi et al., 2007). At the 0.2-0.4 and 0.4-0.6 m depths, Mg 2+ concentrations also decreased linearly, but this was different from the layers above, as no significant difference between the factorial average and the control average was observed. In these layers, incoming Mg 2+ from surface layers leached down due to PG application, compensating in part for the Mg 2+ that migrates downward, especially at the rates of 3 and 6 Mg ha -1 . In the 0.6-0.8 m layer, no effects of rate increase or split applications were observed, however, PG increased the Mg 2+ concentration, since the factorial mean was higher in relation to the control mean, indicating that Mg 2+ was accumulated in this layer (Table 3).
In view of the effect of PG on Mg 2+ availability in the soil, it is important to consider the initial Mg 2+ concentrations in different soil layers before recommending PG rates, because the redistribution of the Mg 2+ concentrated superficially in NT soils may be benefic for crops. It is also necessary to maintain Mg 2+ concentrations above the critical level of the nutrient in surface layers to ensure adequate Mg 2+ supply. Therefore, the planning of PG application must consider other agricultural practices that interact in this context, such as the use of dolomitic lime, magnesium thermophosphate, or other of Mg 2+ sources, depending on the situation.
The soil Ca/Mg ratio increased linearly with PG application rates in all soil layers, although in the 0.4-0.6 and 0.6-0.8 m layers, the increase was not significant in relation to the control (Table 3). In the 0.0-0.1 m layer, the application of single rates (P1) resulted in a lower Ca/Mg ratio compared with rates where the applications were distributed over two (P2) or three (P3) years. Despite not being significantly different, the Ca 2+ concentrations were higher while those of Mg 2+ were lower in P2 and P3, where the time between PG application and soil sampling was shorter than in P1. The combination of these variations resulted in a statistical difference for the Ca/Mg ratio among split treatments. A Ca/Mg increase in response to PG application rates was also reported by Caires et al. (2011a), attributing these results to increases in Ca 2+ concentrations and Mg 2+ leaching.
There was no effect of treatment on soil K + concentrations (Table 4). Despite K + competes for the same adsorption sites as Ca 2+ and Mg 2+ , the downward migration of K + was not comparable to Mg 2+ . This movement for K + (low magnitude) after PG application was reported by Caires et al. (1998Caires et al. ( , 2002, but other studies did not show the same behavior   (2) Linear adjust and R 2 value after the bar; * and **: significant at p<0.05 and p<0.01, respectively; ns: not significant.
Rev Bras Cienc Solo 2016; v40:e0150155 (Raij et al., 1998;Caires et al., 2004). It is possible that K + , added periodically at crop planting throughout the years, compensates any effect possibly caused by PG mobilization. Nevertheless, since the K + concentrations in the soil were considered high, the amount of K + available for leaching was high, therefore, the absence of the effect of PG is evidence that mobilization would be lower for K + than for Mg 2+ , and, or, that the relatively higher cycling of K + (kg ha -1 ) by the crops reduces the leaching effect (Raij et al., 1998). Another hypothesis is that the KSO 4 − formed after the PG reaction can bind to positive charges in the soil, which is not the case for MgSO 0 4 , making it more prone to leaching. Zambrosi et al. (2008) evaluated the ionic speciation of a clayey Oxisol under NT, and found that K + was not complexed by organic anions, and that the complexation with inorganic anions (KCl, KNO 3 and KSO 4 − ) represented a maximum of 0.2 % of the total, remaining as free ions and favoring soil adsorption.
Compared with the control treatment, the application of PG increased the soil S concentration and unlike for the other nutrients, there was an interaction between application rate and splitting for all soil layers ( Table 4). The S concentration increased in the 0.0-0.1, 0.1-0.2 and 0.2-0.4 m layers in the following order: P1<P2<P3. As PG rates increased, the S concentrations also increased linearly, in all splitting treatments and soil layers, except for the 0.0-0.1 m layer where this response occurred only in P2 and P3. The effect of split applications changed in the deeper layers, so that in 0.4-0.6 m, no difference was observed between P1 and P2 and both were exceeded by P3, while in the 0.6-0.8 m layer, no splitting effect occurred. Increased S concentration in the soil was also reported in other studies, to a depth of 0.6 m (Caires et al., 2011a) and 0.8 m (Raij et al., 1998). In addition, the above results indicate that splitting the PG application could benefit the S levels in the soil, especially in the near-surface layers, improving S cycling by plants and increasing the soil stock.
The retention of SO 4 2− in the soil is low, especially in the topsoil where electronegativity is higher. In NT, soil net charge is even more negative in this layer due to the accumulation of organic matter and surface liming (Casagrande et al., 2003), thus repelling SO 4 2− . In fact, due to the absence of tillage and to the natural low mobility of phosphate anions, they accumulate in the topsoil along with fertilization in the furrow during crop planting, and phosphates are preferably retained in the anion exchange capacity (AEC). Higher concentrations of S are common in subsurface layers (Borges, 1997;Alvarez V et al., 2000), where lower organic matter content allows for a higher AEC (Lima et al., 2013). In this study, this trend was estimated by the ∆pH values, which were less negative for the 0.4-0.6 and 0.6-0.8 m layers, -0.92 and -0.80, respectively, in comparison with the values for the 0.0-0.1, 0.1-0.2 and 0.2-0.4 m layers, where the ∆pH values were -1.13, -1.04 and -1.04, respectively.
Phosphorus concentrations in the soil were not affected by treatments (Table 4), although residual concentrations of P (0.2-0.6 % P 2 O 5 ) are present in PG, due to its origin in the phosphate fertilizer manufacturing process. In this study, the combination of the broadcast application of PG to the area and the very clayey soil texture creates a large P retention capacity, preventing increases in available P concentrations in the soil analysis.

Macronutrient plant leaf tissue concentrations
The N concentration of corn, wheat and soybean leaves (Table 5) were not affected by any of the treatments. Raij et al. (1998) and Caires et al. (2011a) found similar results regarding corn, and despite the fact that rainfall in December 2011 was about 100 mm lower than the monthly mean, there was no water stress in the corn growing season (Figure 1), with regular precipitation every 7-10 days. Wheat emerged on July 30 th , 2012 and was affected by water stress in the early stages of its life cycle, with rainfall precipitation in August being only 2 mm while there was 100 mm less precipitation than the monthly mean of September. In this case, the absence of a Rev Bras Cienc Solo 2016; v40:e0150155 possible effect of PG on leaf N concentrations could be due to absorption limitations imposed by the dry period, which reduced wheat growth and the yield. No effect of PG application on wheat leaf N concentration under water stress was also reported by Caires et al. (2002). Soybean obtains most of the necessary N through symbiosis, resulting in a reduced potential effect of PG on N absorption from the soil, as was observed in this study and also reported by Nogueira and Melo (2003) and Caires et al. (2003;2011a).
The treatments had no effect on leaf P concentrations (Table 5), corroborating findings of other studies for corn (Raij et al., 1998;Caires et al., 2004), wheat (Caires et al., 2002), and soybean (Nogueira and Melo, 2003). The same concordance between the soil and leaf tissue results occurred for K in corn, wheat and soybean (Table 5). Despite increasing Ca 2+ concentrations in soil layers due to PG application, K plant uptake was not influenced by competitive inhibition. In this case, the high concentrations of K + in the soil and K + supply through NPK fertilization in planting furrows for each crop may explain the results. Other studies with PG also reported similar results for leaf K regarding corn (Caires et al., 2011a), wheat (Caires et al., 2002) and soybean (Caires et al., 2003(Caires et al., , 2011aNogueira and Melo, 2003).
The reduction in Al 3+ concentration combined with Ca 2+ increase throughout the soil profile improved soil fertility, especially in the subsurface, which may have benefited root growth (Carvalho and Raij, 1997) and enhanced root interception, the second most important ion-root contact component for Ca absorption (Prado, 2008). Thus, the linear increase in leaf Ca concentration for corn, wheat and soybean (Table 5) in response to PG application rates was explained. Increases in leaf Ca concentrations were also reported in other studies for corn (Caires et al., 2004;2011a) and wheat (Caires et al., 1999;2002), as well as soybeans (Caires et al., 2003(Caires et al., , 2006Rampim et al., 2011).
The decrease in Mg 2+ concentrations in soil layers to a depth of 0.60 m was followed by a reduction in Mg absorption by corn, wheat and soybean (Table 5). High soil concentrations of Ca 2+ and K + inhibit Mg absorption by ionic competition, leading to possible nutrient deficiency (Prado, 2008). However, wheat was the only crop with Mg leaf concentration below the sufficiency range (1.5 -4 g kg -1 ), and only for 6, 9 and 12 Mg ha -1 of PG. A decrease in Mg concentration in corn leaf tissue after PG application was reported also by Raij et al. (1998) and Caires et al. (2004). For wheat, Caires et al. (2002) reported no effect of PG application on Mg leaf concentration, while Rampim et al. (2011) also observed a decline in leaf Mg, and the studies of Caires et al. (2003; and Gelain et al. (2011) found increasing PG application rates related to a decrease in Mg concentration in soybean leaves.
The S concentrations in corn, wheat and soybean leaves (Table 5) were linearly increased by PG application rates, reflecting the S increase in the soil due to PG application rates. This result was similar to those reported in other studies on corn (Caires et al., 2011a), wheat (Caires et al., 2002) and soybean (Caires et al., 2003(Caires et al., , 2011aNogueira and Melo, 2003). In contrast, although application splitting also resulted in a higher soil S concentration to 0.6 m deep, this factor did not affect leaf S concentration in any of the crops; the same was true for the Ca and Mg leaf concentrations, even though splitting resulted in a higher Ca/Mg ratio in the 0.0-0.1 m soil layer. The macronutrients did not demonstrate a response in terms of leaf concentration for any of the studied crops due to the effect of splitting, demonstrating that the magnitude of the variation in soil properties due to splitting the PG application rates was not high enough to affect plant uptake.

Crop yields
Corn yield was not influenced by splitting, but the effect of PG application rate was quadratic (Figure 2), with a decrease at higher rates, especially at 12 Mg ha -1 of PG,   (2) Linear adjust and R 2 value after the bar; * and **: significant at p<0.05 and p<0.01, respectively; ns: not significant.  which produced 0.3 % less than the control. The maximum technical efficiency (MTE) rate of PG was 6.38 Mg ha -1 , corresponding to 10.94 Mg ha -1 of corn, which is about 5.5 % higher than the control yield. Similar results in another study with PG application rates in Guarapuava, with a quadratic effect and a MTE of 7.8 Mg ha -1 of PG was reported by Caires et al. (2011a), concluding that the yield decrease at higher rates was due to Mg 2+ and K + leaching from the topsoil.
Leaf P, K, Ca and Mg concentrations were within the sufficiency ranges, whereas N and S concentrations were above the sufficiency range for corn (Santos et al., 2009). Although the Mg concentrations were higher than the critical level for the crop, the average with 12 Mg ha -1 of PG was 23 % lower compared with the control, consistent with the lower yield. Unlike corn, wheat yield increased linearly with PG rates, without any effect of split applications (Figure 2), and the significance in the contrast between control and factorial treatments showed that by the average of the rates and splitting the application of PG improved wheat yield.
Soybean showed no yield response to treatments and no significant difference between the control and the factorial treatments ( Figure 2). Leaf concentrations of macronutrients were within the sufficiency ranges for all treatments, except for 12 Mg ha -1 of PG in the case of N and for all rates in the case of S, where concentrations exceeded the sufficiency range for the crop (santos et al., 2009). No effect of PG application on soybean yield was reported by Caires et al. (2003;2011b), Neis et al. (2010) and Rampim et al. (2011). The roots of this crop, compared with corn for example, have a higher root cation exchange capacity (Fernandes and Souza, 2006), leading to a higher efficiency in the accumulation of soil divalent cations, such as Ca 2+ and Mg 2+ , in the rhizosphere, which favors absorption even under low concentrations of these cations in the soil. Soybean roots are influenced slightly by low Al 3+ concentrations in the soil when rainfall is regular (Caires et al., 2001), as was the case in this study.

CONCLUSIONS
The use of phosphogypsum increased grain yield of poaceous crops, with a quadratic response to the rates for corn under a normal growing season, and a linear response for wheat in a growing season with water restriction. Soybean yield was not significantly influenced.