Safflower seeds were hand-planted at a depth of 3 to 5 cm with spacing of 0.45 m and distance of 0.10 m in each row. Fertilization and other crop management practices were carried out manually. The seeds were treated with Tiram-based fungicide and sown directly into the ground (thirty seeds per meter). No fertilizers were applied during planting. Safflower seeds were sown on May 22, 2014 in conventional tillage system.

The experimental design was completely randomized in a 13 x 2 factorial design with four replications. The treatments consisted of 13 increasing nitrogen rates (0, 50, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550 and 600 kg ha^-1) under two water regimes: irrigation and rainfed conditions. Each plot consisted of three 6-meter rows. Analyses were carried out on the middle row. Two applications of nitrogen were performed per week, starting at 30 days after sowing (DAS) and ending at 75 DAS, when flowering began.

Safflower plants were irrigated in order to meet the local average atmospheric demand, 3.5 mm day^-1, through a drip tube system distributed along the crop rows. Irrigation treatments were applied when the plants presented 50% senescent leaves. The treatments under rainfed conditions only received rain water. Thus, there was a supply of 315 mm more water in the four-month cycle in the treatment under irrigation conditions, totaling 466.4 mm. Under rainfed conditions there was only precipitation of 151.4 mm (Figure 1).

Figure 1

The variables plant height measured from ground level to plant apex, number of heads per plant determined by the harvest of four random plants from the middle rows were evaluated before the harvest during the physiological stage of maturity.

Grain mass per plant, 100-seed weight obtained by the mass of one hundred grains collected from four random samples, and productivity were determined during the harvest at September 30, 2014. The grain yield in each plot was adjusted to 10% moisture.

Oil content was determined by TD-NMR analysis in an SLK-SG-200 spectrometer (SpinLock Magnetic Resonance Solutions) at 25°C equipped with a permanent magnet of 0.23 T (9 MHz for 1H) and a probe with 13 mm x 30 mm useful area. Software used consisted of Condor IDE with CPMG pulse sequence and Qdamper, with results being expressed on a dry basis (% DB).

Gama transformation was used for data that did not follow normal distribution. Analysis of variance and Tukey's mean comparison test at 5% probability were applied to the sources of variation, using Minitab 16 statistical software(r). As for the source of variation of nitrogen levels, regression analysis was performed with plateau function (broken line). It was considered ready for presentation when the fit was significant at 5% probability and R² was greater than 0.70.

Safflower plant height was affected by the interaction among factors. Intermediate rates of nitrogen (250 - 400 kg ha^-1) applied to treatment under irrigation conditions provided benefits (Table 1). Maximum safflower plant height based on the linear-plateau model was 120 cm (350 kg ha^-1) under irrigation conditions and 113 cm (199 kg ha^-1) under rainfed conditions (Figure 2A).

Table 1

*Means followed by different lowercase letters in the line and capital in column indicate statistical difference byTukeytest at 1% probability.

Figure 2

The number of heads per plant was also influenced by nitrogen and irrigation conditions, and also by the interaction among factors (Table 1). However, regression analysis did not indicate any polynomial models. High rates of nitrogen reduced the number of heads per plant regardless of the water regime. In contrast, low rates of nitrogen were dependent on irrigation to increase the number of heads per plant. Thus, less nitrogen was needed under irrigation conditions to increase the number of heads per plant than under rainfed conditions.

In this study, the 100-seed weight was dependent on the interaction among factors. The rate of 600 kg ha^-1 in the rainfed treatment provided the largest 100-seed weight (5.05 g) (Table 1). The optimum point was not determined since R² was inferior to 0.70 (data not shown). Grain yield per plant was significantly affected by the interaction among factors. The rate of200 kg.N ha^-1 applied to the treatment under irrigation conditions resulted in 49.67 g of grains per plant (Table 1).

The interaction among factors affected grain yield (Table 2), which requires a maximum rate of 208 kg N.ha^-1 to present gains (Figure 2B) according to the linear-plateau model.

Table 2

*Means followed by different lowercase letters in the line and capital in column indicate statistical difference byTukeytest at 1% probability.

This rate results in a maximum grain yield of 4005 kg.ha^-1 under irrigation conditions. In what concerns to rainfed conditions, the model proposed that the rate of 214 kg N.ha^-1 provides a maximum of 2000 kg of grains ha^-1 (Figure 2B).

Safflower plants grown under rainfed conditions and treated with 300 kg N.ha^-1 presented the highest oil content (Table 2) according to the interaction and linear-plateau model (Figure 2C). Under irrigation conditions, the model demonstrated that 400 kg N.ha^-1 would be necessary to achieve maximum oil content, just as observed in the interaction (Table 2).

Results obtained in this study for plant height match those found by Aghamohammadreza, Mirhadi, Delkhosh & Omidi (2013), who observed heights from 75 to 150 cm under different crop conditions. Siddiqui & Oad (2006), in Pakistan, observed that the maximum rate of 180 kg ha^-1 led to higher plant height (165 cm). However, the ineffectiveness of low nitrogen rates (0, 40, 80 kg.ha^-1) on safflower growth was observed by Elfadl, Reinbrecht, Frick & Claupein (2009), in marginal areas of south-west Germany.

According to Bagheri & Sam- Dailiri (2011), drought may cause reduction on plant height. Thus, irrigated treatments respond better to nitrogen fertilizers, which is certainly related to increasing recovery and dilution of nitrogen in the soil.

Ebrahimian & Soleymani (2013), achieved a maximum number of heads per plant with the application of 150 kg N ha^-1. Dordas & Sioulas (2008) and Rastgou et al. (2013), report that nitrogen fertilization increased the number of heads per plant by 32% and 31%, respectively. In contrast, Strasil & Vorlicek (2002), observed that nitrogen fertilization did not affect the number of heads per plant under rainfed conditions.

Vafaie, Ebadi, Rastgou & Moghadam (2013),observed variation between 5.07 and 5.61 in the 100-seed weight. In safflower cultivation under rainfed conditions with nitrogen application ranging from 0 kg ha^-1 to 150 kg ha^-1. Similarly, Rastgou et al. (2013), reported a 30% increase in the 100-seed weight with the application of 150 kg N.ha^-1. Strasil & Vorlicek (2002), report that nitrogen fertilization did not affect the 100-seed weight.

The rainfed treatment required 250 to 400 kg N.ha^-1 to show an increase in grain yield per plant, demonstrating better efficiency of nitrogen fertilization with the use of irrigation, which is mainly related to fertilization processes and grain filling per plant. Higher grain yield per plant was also observed by Dordas & Sioulas (2008), with the application of nitrogen.

Miri & Bagheri (2013), also observed low er productivity under rainfed conditions (463 kg.ha^-1), whereas using an irrigation system led to 3845 kg ha^-1. Hasanvandi, Aynehband, Rafiee, Mojadam & Rasekh (2014), observed an increase in safflower grain yield of 33% and 25% in two years of cultivation in Iran.

Dordas & Sioulas (2008), reported that ni trogen fertilization increased the production of safflower grains by, on average, 19%. These authors achieved maximum yield under rainfed conditions with the application of 150 kg N. ha^-1. Strasil & Vorlicek (2002), also achieved maximum safflower grain yield with the application of 150 kg N ha^-1. Rastigou et al. (2013), report positive effects of nitrogen fertilization on safflower grain yield, reaching higher yields at a rate of 200 kg.N ha^-1. Gholinezhad, Aynaband, Ghorthapeh, Noor-mohamadi & Bernousi (2009), in a study on saf-flower cultivation under different water regimes and nitrogen levels in Iran, report that the ideal conditions for safflower cultivation are irrigation at 90% field capacity and 220 kg N.ha^-1. Recently, Sampaio et al. (2016), reported that the maximum yield of safflower seeds was achieved with up to 260 kg N.ha^-1 with sowing in autumn in Brazil.

Sampaio et al. (2016), achieved maximum oil yield of safflower under rainfed conditions in Brazil when 244 kg N ha^-1 were applied during sowing. Hasanvandi et al. (2014), observed an in crease in oil content under irrigation conditions, however, the authors point out that this effect is not indispensable for oil content, but for grain yield, just as observed in this study. Dordas & Sioulas (2008), found no increase in oil content with a maximum rate of 200 kg N ha^-1.

Rastigou et al. (2013), report that there was no increase in oil content with the application of nitrogen fertilizer, however, rates above 200 kg.N ha^-1 caused reduction in oil content, just as observed in the present work, but with higher rates. According to Sabbagh et al. (2012), increas ing nitrogen fertilization rates too much leads to increased nitrogen absorption, resulting in the production of amino acids and other compounds in peptide bands. Thus, increased amino acid production causes low oil content.

Safflower oil yield was benefited by rates from 200 to 350 kg N.ha^-1 according to the interaction and by irrigation conditions (Table 2), just as observed by Miri & Bagheri (2013) and Hasanvandi et al. (2014). Thus, the fit determined that 200 kg N ha-1 are enough to provide an oil yield apex, with 1200 and 600 kg.ha^-1 under irrigation and rainfed conditions, respectively (Figure 2D). Despite oil content being less affected by nitrogen fertilization, an increasing in grain yield compen sated and increased this variable. Dordas & Siou-las (2008), observed an increasing of 21% with the application of nitrogen, being this fact due to grain yield. It is evident that nitrogen affects more sharply safflower grain yield than its oil content, just as stated by Hasanvandi et al. (2014).

The application of nitrogen has proved to be particularly effective on safflower yield in various regions (Eryigit, Aki§ & Kaya, 2015). Determining the appropriate rate of nitrogen fertilizer provides increased production of safflower grains, mainly under the Brazilian tropical conditions due to severe decomposition rates and nitrogen dyna mics in the soil.