Response of wheat (Triticum aestivum L.) to sulfur impurity supplied as triple-super phosphate under gypsum and urea fertilizer backgrounds

On-farm experiments (24) were conducted in two seasons to assess the influence of sulfur impurity in triple-super phosphate (TSP) on wheat. The experiments were laid-out in completely randomized block design in triplicate. Nutrients investigated include: Nitrogen (N), sulfur (S) and phosphorus (P). The TSP is tested to contain about 2 to 6% sulfur as impurity. However, wheat did not show response to sulfur from TSP. Though, it was not statistically significant, however, there were always yield increments by certain percent due to S-TSP in eight out of 10 sites with increasing trends of yield curves. Overall, the effects of such nutrients incidentally supplied from concentrated fertilizers like TSP should not be overlooked, as the benefits could be expressed in quality attributes of crops. Indeed, this is vital in varietal specific nutrient requirement studies. Overall, the benefits of such small-dose of nutrients could be many-folds to small-holders if integrated with organic resources, thereby encouraging organic agriculture. But, wheat showed response to sulfur from gypsum (in 67%) and N from urea (100%) of sites. As depicted, the sharply-rising yield curves with applied nitrogen elucidate that nitrogen was the most limiting nutrient followed by S. Always rising yield response curves also show strong positive synergies between the applied nutrients.


INTRODUCTION
Soil fertility decline in Ethiopia has been well documented with most attention that has been given to nitrogen (N), phosphorus (P) and potassium (K). Sulfur (S) is a macronutrient that is taken-up by grain crops in amounts similar to those of P, 10 to 30 kg/ha (Scherer, 2001;Jamal et al., 2010;Mengel and Kirkby, 1987). According to Weil (2011) adequate level of sulfur in soils is very important both for the satisfactory growth of plants; and also for ensuring optimum levels of S-containing essential aminoacids, oils, vitamins and flavored compounds in plants. The essential amino-acids methionine (21% S), cysteine (26% S) and cystine (27% S), which are the building blocks of proteins in food and feed contain S (Chattopaddhyay and Ghosh, 2012). It is also known to enhance other nutrients use efficiency; and ranks second only to N in importance for optimum crop yield and quality produce (Brown et al., 2005). For example, wheat protein is rich in non-essential glutamic-acid and proline, whereas deficient in most essential-amino acids such as lysine, tryptophan, threonine, methionine and histidine (Khan et al., 2014). These amino-acids contain sulfur, and therefore, without modest or significant supply of sulfur, crop-plants can neither express their full genetic potential yield/quality nor can complete their life cycle (Khan et al., 2014). But, when planning the nutritional requirement it is well recognized that oil crops and legumes have high sulfur demand. Whereas, cereals have lower sulfur demand, and are reported to remove about 10 to 20 kg S/ha for producing the grain yields of 8 Mg/ha (Walker and Booth, 1992;McGrath et al., 1996;Oates and Kamprath, 1985). Similarly, Zörb et al. (2013) reported a modest amount of S, 15 to 35 kg S/ha for better quality and optimum wheat yields.
With respect to its response, crops like wheat can respond to sulfur rating between 5 and 10 kg S/ha (Weil, 2011). A study by Menna et al. (2015Menna et al. ( , 2016 reported 5 to over 20 kg S/ha or even less for wheat, depending on soils. Indeed, this is an amount of sulfur that can be supplied to plants from accidental applications of concentrated fertilizers like TSP and DAP. These highanalysis fertilizers, however, contain significant amount of S, 2 to 6% by weight (Weil, 2011;Weil and Mughogho, 2000). This is a micro-dose amount that can significantly increase small-holding farmers' yields in precisionfarming. In the view of the aforementioned, therefore, this paper sought to (1) evaluate wheat response to the incidental application of S impurity from TSP; (2) to infer the most yield limiting nutrient from yield response curve(s); and (3), to see the interaction effect of S, N and P following wheat yield trends.

Site selection
On-farm field experiments were conducted in three representative locations, namely Arsi (Ar), East Shewa (ES) and West Shewa (WS) zones in central Ethiopian agricultural lands. The three representative locations and salient features are presented in Table  1.

Experimental materials, treatments and design
Gypsum samples were collected from six curie sites and analyzed for the SO 4 -S contents and/or for its purity. The materials were tested to contain the SO 4 -S contents of 13.5 to 18.0%. Then the samples with 18.0% SO 4 -S were taken as experimental material. In order to investigate the response of wheat to S from TSP, 24 onfarm experiments were conducted in two seasons. The sites were geo-referenced using Global Positioning System (GPS) GARMINmodel # GPS-60 assisted by Google Earth (2011) as presented in Table 1. In season-I (2013/2014), 18 experiments were conducted, that is, 6 per zone (location) covering different agro-ecological zones (AEZs). In this season, four treatments: absolute control (CK); N alone = N 1 ; nitrogen + sulfur or NS = N 1 S 1 ; and nitrogen + phosphorus + sulfur or NPS = N 1 P 1 S 1 were tested. The nutrients evaluated were 2-levels of S (0 and 20 kg/ha), 2-levels of P (0 and 20 kg/ha), and 2-levels N (0 and 69 kg/ha).
In season-II (2015/2016), another 6 experiments were conducted (2 per zone). In this season, three study sites, namely Gora Silingo 2 (GS 2 ), Keteba 2 (Ke 2 ), and Nano Suba 2 (NS 2 ) were selected because of the wheat response to S from gypsum in season-I. Whereas, Wonji Gora1 (Do 2 ), Bekejo 2 (Bk 2 ) and Berfeta Tokofa 2 (BT 2 ) were selected randomly without pre-soil testing, but on areas some 0.5 to 1.5 miles away from last season S responsive sites, namely Do 1 , Bk 1 and (BT 1 ), respectively. In this set, 9 treatments were tested: CK; nitrogen alone or N = N 1 ; nitrogen + sulfur or NS 1 ; nitrogen + sulfur or NS 2 ; nitrogen + sulfur or NS 3 ; nitrogen + phosphorus or NP = N 1 P 1 ; nitrogen + phosphorus + sulfur or NPS 1 ; nitrogen + phosphorus + sulfur or NPS 2 ; and nitrogen + phosphorus + sulfur or NPS 3 . Nutrient evaluated were 4-levels of sulfur (S) (S 0 = CK, S 1 = 5, S 2 = 10 and S 3 = 20 kg S/ha); 2-levels of nitrogen (N) (N 0 = CK, N = 69 kg N/ha); and 2-levels of phosphorus (P) (P 0 = CK and P 1 = 20 kg P/ha). In both seasons, randomized complete block design (RCBD) was used in triplicate. In both season, each replicate had 12 plots, with an area of 3 m × 5 m = 15 m 2 . Agronomic spacing for wheat (25 × 5 cm between rows and plants, respectively) was used. Each plot had 12-rows of plants with 2 borders in each side and another one-row for plant sampling. The central rows with an area of 4 m × 1.5 m = 6 m 2 ) were used for agronomic data collection. Nitrogen was split applied; where 1/3 was incorporated into soils before seeding and the remaining 2/3 was top-dressed at the stage of tillering. Whereas, the entire sources of S and P were incorporated into soils just before seeding. A wheat cultivar known as "Kekeba" was used as a test-crop.  Bouyoucos (1962) Saturation percent (SP); total nitrogen (TN); organic carbon (OC); available phosphorus (P); sulfate sulfur (SO 4 -S); electrical conductivity (EC); exchangeable (Exch.); and cation exchange capacity (CEC); na = not applicable.

Soil sampling, preparation and analysis
Soil samples (24) were collected in two seasons or sets. The first set, 18 samples were collected in season-I before planting each of the 18 gypsum (S + Ca) response experiments. The second set, 6 samples were collected in season-II before planting each of the 6 S rate experiments. In doing so, soil samples representing each block were taken from 10-spots (0-20 cm soil depth) and bulked together. Then the samples were further composted to make one sample per farmer field and air dried in dust free rooms. The dried samples were ground and made to pass through < 2 mm sieve and analyzed for the variables as per the methods shown in Table 2.

Agronomic data analysis
Yield data were analyzed using SAS Version-9 (SAS Inst. Inc. 2012). The ANOVA was done using PROC-MIXED in SAS protocols to evaluate treatment differences. When the differences between treatments were significant, the least significant difference (LSD) was used to separate means at 5%, 1% or 0.1% probability levels. More specifically, pairwise orthogonal comparisons among treatments using SAS contrast statements were made to determine the significance of treatments at each level or to determine the effect of S as an impurity from TSP (S-TSP) together with N-urea and P-TSP on wheat yield. Table 3 presents the properties of initial soils sampled before planning in the two seasons. As shown, in WS zone the pH ranged from strongly to moderately acidic, whilst it was strongly acidic to near neutral in Arsi zone. In ES zone it ranged from neutral to moderately alkaline (calcareous). The nitrogen (N) content is very low or low based on the ratings by Landon (1991). Available P at the ES and Arsi zones was either very low or low as per the ratings by (Horneck et al., 2011). In such low P soils, fertilizer responses are most likely expected, but in the calcareous soils of ES and in strongly acidic soils of WS its availability can be limited due to the precipitation reactions with Ca and aluminium (Al). With respect to sulfate sulfur (SO 4 -S), based on the ratings by Tandon (1991) over 50% of soils were found to be S limiting. Organic carbon (OC) content of the soils in the first season was also either very low or low. In only about 30% of soils the OC were either marginal of above the threshold values.

Response of wheat to S impurity in TSP
Ethiopia needs to increase agricultural production in order to feed an ever increasing population estimated at 102 million in 2017 (EEA/EEPRI, 2017). However, to ensure cost-effective and quality produce, healthy soils are needed. Sulfur even in its micro-dose level is essential not only for plant growth and quality produce, but also enhances other nutrients' use efficiency and ranks second only to N in importance for optimum crop performance (Brown et al., 2005;Zörb et al., 2013). The following discuss wheat response specifically to S impurity from TSP (S + P) in relation to gypsum (S + Ca) and urea (N). For this purpose, sites/soils which did not show response to P from TSP or P-sufficient sites are considered.
Responses and/or benefits of sulfur are easily overlooked where a basal dressing of P is applied as inorganic triple-super phosphate (TSP), a highanalysis fertilizer which is commonly assumed to be free of S (Shenkalwa, 1986). However, TSP typically contains agronomic-ally significant quantities of S, 2 to 6% by weight (Weil, 2011;Weil and Mughogho, 2000). According to the study, TSP is a soluble fertilizer primarily containing about 20% total P (44-48% P 2 O 5 ) and 13-15% calcium (Ca). It is also estimated to contain a maximum amount of 4% residual phosphoric acid (H 3 PO 4 ). Sulfur is recognized to be associated with TSP during its manufacturing from rock phosphate and sulfuric acid (H 2 SO 4 ). To see such effects on crop yields, pre-planned pair-wise orthogonal comparisons among treatments using SAS contrast statements were done on soils that showed response to applied S from gypsum (significantly or marginally), but not to P. The major target sites for such analysis were: AA 1 , Do 1 , BL 1 , GS 2 (Ar zone); Ud 1 , Ki 1 , Ke 2 and Bk 2 (ES); and NK 1 /(DL 1 ) (WS) in seasons-I and II (Tables 4 to 8). In this assessment, the effects of S-TSP can be seen from the yield gap between treatments NPS 3 and NS 3 . It is well noted that wheat cultivar showed significant responses (p ≤ 0.001) to S-gypsum on average, in 67% of sites. However, the responses of wheat to S-TSP were not statistically significant, or the ANOVA did not reveal such responses. This might be due to the inherently low levels of available P in soils, which can be seen by looking  at yield responses due to P, that is, the yield gaps due to the treatment effects between NP and N. But, the comparisons showed non-significant negative limits at 95.0% (Tables 4 to 8 and Figures 1 to 10), indicating that there were no wheat grain yield responses to S from TSP (that is, beyond 20 kg S/ha, the amount of S supplied as mineral gypsum). If a site is responsive to S-gypsum, but not to P-TSP, then any yield increase beyond 20 kg S/ha can be regarded as the response coming due to S-TSP when keeping other factors constant. The major reason could be that the response of wheat to S that is expected from TSP might have been obscured due to the inherently low levels of P in soils (Table 3). This can easily been seen by looking at the yield gaps between NS 2 and NS 3 (responsive = r); NS 3 and NP (nonresponsive = ns); and NP and NPS (responsive = r) in the x-axis of Figures 1 to 10 and Tables 4 to 8. Then, the observed difference in the first two and last two, and the lack of response between the second two treatments may suggest that the amount of S from TSP (that is, 2 to 6 kg S/ha) also might not be sufficient enough to bring the intended statistically significant yield. This can also be affirmed by looking at the lower treatments, for example (at 5 kg S/ha). This is to mean that there was yield increase due to S from gypsum, but which was not statistically significant. Though, the responses of wheat to S from TSP were not statistically significant, there are always yield increments between NS 3 and NP (nr) treatments by certain percentages. So, such yield progressions should not be overlooked, because it can be expressed in the quality attributes of crops like wheat. For example, at GS 2 site there was grain yield response due to N and S at all levels, except for S that is expected from TSP (that is, the yield gap between NPS 3 -NS 3 ). But, the yield increments at this point can also be due to P as there was P response in this site, which is obtained as a yield difference between treatments, N and NP. As the initial soils tested low in P, the suggested amount of S-TSP (2-6%) may also not be adequate enough to bring statically significant yield. This reason can be noticeable from yield differences between the lower treatments (N and NS 1 ) as there was no yield response due to S-gypsum at 5 kg S/ha. However, at GS 2 site, S treatments above 5 kg/ha had significant yield increases throughout, suggesting the existence of positive synergy between S and N or P. In general, it should be noted that though not statistically significant, this can be expressed in quality attributes of harvested produces. Therefore, accounting for such nutrient contents existing as impurities in concentered fertilizers like TSP and DAP will be very important in precisionfarming. This is also helpful in avoiding unnecessary fertilizer blending, if used in integration with locally available materials like gypsum. This at the same time will encourage organic farming practices.

Response of wheat to N, S from gypsum and phosphorus from TSP
As graphically presented, wheat showed significant responses to S from gypsum (S + Ca) (Figures 1 to 10). Wheat specifically showed highly significant response (p ≤ 0.001) to N-urea in all sites, but with less response to P when compared with S and N. Sites which did not show response to P in relation to next lower level treatment (that is, NS) are AA 1 , Do 1 , and BL 1 (Ar zone) making 16.7%; and Ud 1 and Ki 1 (ES zone), which make up 5.6% of the sites. Sites with marginal response to P were Do 1 , Ud 1 , DL 1 , and NK 1 which make up 22% of the sites. With respect to S, about 50% sites showed highly significant response (p ≤ 0.001) to S-gypsum as related to soil-test levels; and 22% of the sites had marginal response (p ≤ 0.05). Figures 7 to 10 depict wheat response to S-gypsum, N and/or P at GS 2 , Ke 2 , Bk 2 , and N/S 2 sites in season-II. These are sites which showed Particularly, wheat showed highly significant response *** to N and responses to S were either at ** or *. Most interestingly, in all investigated sites the responses to N, S and/or P are related to soil-test values. From overall results it is observed that, ES had better consistency of wheat response to S at all treatment levels, correlating better with soil-test values compared with Ar and WS zones. Furthermore, in ES zone there was better synergy between S, N or P which is manifested by its always increasing yield advantage with the type and level of nutrients supplied. However, the interaction effect of S with N is more pronounced than that of S. Crista et al. (2013) made a similar observation. Wheat grain yield at AA1 in response to the applied S, N and P (season-I). Means bearing the same letter(s) on bars within a field-site are not significantly different at *p ≤ 0.05 prob. level; significant at **p ≤ 0.01; highly significant at ***p ≤ 0.001; and ns = not significant. % grain yield increase due to S-TSP at: AA 1 = 1.5%; Do 1 = 20.0%. Error bars show the standard error of the mean for each treatment. Source: Author Survey Figure 2. Wheat grain yield at Do1 (Arsi zone) in response to the applied S, N and P (season-I). Means bearing the same letter(s) on bars within a fieldsite are not significantly different at *p ≤ 0.05 prob. Level; significant at (**p ≤ 0.01); highly significant at ***p ≤ 0.001; and ns = not significant. % grain yield increase due to S-TSP at: AA 1 = 1.5%; Do 1 = 20.0%. Error bars show the standard error of the mean for each treatment. Source: Author Survey  According to the authors, N and S have a highly significant influence and contribution to protein contents of crop produces than P. Conversely, P has no such greater influence on protein content other than supporting the assimilation and metabolism of absorbed N forms (Crist, et al., 2013).
As depicted in the figures, in all sites with the applied N there were sharp-rises in yield curves even including the P and S non-responsive sites like Boneya Edo (BE 1 ) indicating that N was the most yield limiting element Figure 5. Wheat grain yield response at Ki to S, N, and P (season-I). Means bearing the same letter(s) on bars within a site are not significantly different at p ≤ 0.05 prob. level. Percentage grain yield increase due to S from TSP at: BL 1 = 12.0%, Ud 1 = 12.2% and Ki 1 = 6.5%; NK 1 = 13.8%, and/or similar DL 1 = 13.9%. Source: Author Survey Figure 6. Wheat grain yield response at NK/DL to S, N, and P (season-I). Means bearing the same letter(s) on bars within a site are not significantly different at p ≤ 0.05 prob. level. Percentage grain yield increase due to S from TSP at: BL 1 = 12.0%, Ud 1 = 12.2% and Ki 1 = 6.5%; NK 1 = 13.8%, and/or similar DL 1 = 13.9%. Source: Author Survey followed by S and/or P.
With applied N in in terms of yield gains, for example, 72.2 to 148.7% grain yield advantages over control in season-II alone were obtained. This further affirms that N was the most limiting element in the studied soils. So, supplying soils first with adequate amounts of N is of paramount importance to get responses or benefits from any other kind of added essential element(s). Indeed, this is strongly linked with the low contents of OM vis-à-vis the dynamics of N in tropical soils. Wheat grain yield at GS 2 site in response to N, S and P nutrition (season-II). Means bearing the same letter(s) within a group are not significantly different statistically at the probability level analyzed by t-test. Percentage grain yield increase due to S from TSP at: GS 2 = 1.23%; and Ke 2 = 4.64%. Source: Author Survey Figure 8. Wheat grain yield at Ke 2 site in response to N, S and P nutrition (season-II). Means bearing the same letter(s) within a group are not significantly different statistically at the probability level analyzed by t-test. Percentage grain yield increase due to S from TSP at: GS 2 = 1.23%; and Ke 2 = 4.64%. Source: Author Survey Overall, the maximum grain yield recorded in the present experiments was only about 6.6 t/ha. The average grain yield of wheat, however, can reach ≥ 8.5 t/ha under optimal conditions (Zhao et al., 1999). Indeed this indicates the significance of multitude of factors analysis that is expected to be affecting crop yields. Interestingly, all sites had positive yield gains by certain percentages, except, Bekejo 2 and Nano Suba 2 sites. These two sites had slightly negative yield disadvantage in the treatments between NP and NS 3 . For example, at Bk 2 site the grain Figure 9. Wheat grain yield response to applied N, S and P nutrients at Bk 2 site (season-II). Means bearing the same letter(s) within a group are not significantly different statistically at the probability level analyzed by t-test. Percentage grain yield increase due to S from TSP at: Bk 2 = -3.08%; N/S 2 = -2.87%. Source: Author Survey Figure 10. Wheat grain yield response to applied N, S and P nutrients at N/S 2 site (season-II). Means bearing the same letter(s) within a group are not significantly different statistically at the probability level analyzed by t-test. Percentage grain yield increase due to S from TSP at: Bk 2 = -3.08%; N/S 2 = -2.87%. Source: Author Survey yield obtained from NS 3 treatment was 3.9 t/ha, slightly greater than that from treatment NP (3.8 t/ha). At N/Suba 2 site also the yield obtained from treatment NS 3 (4.18 t/ha) was greater than that from treatment NP (4.06 t/ha). This may indicate that in highly S deficient soils, the testcrop's tendency in absorbing more S than P. Indeed, the soils from these sites were tested very low in SO 4 -S compared with others, showing better interaction of N with S in impacting wheat yield under S limiting soils than with P. Such strong interaction effect of N with S on wheat yield and quality attributes was also reported by other workers (Habtegebrial and Singh, 2009;Reussi et al., 2012;Saeed et al., 2013).

CONCLUSIONS AND RECOMMENDATIONS
From the results it can be concluded that there is wheat yield incensements by certain percentages with applied S impurity from TSP in 8 out of 10 target sites, but not statistically significant. These percentages are indicated as captions under the figures. Such small percentage yield increments, however, can be big enough to be expressed in quality attributes of harvested produces. And, this small-dose shouldn't be overlooked as crop varieties differ in their nutrient requirements. This microdose amount of nutrient will be important not only in terms of cost-effective fertilizer managed farming practices, but also important in terms of environmental concerns. It should further be noted that such small amounts of S are important to sustain crop production, if integrated particularly with organic resources, because smallholding farmers can afford to buy, 5 to 10 kg bag of similar fertilizers, thereby encouraging organic agriculture. With respect to N, it is noted that in all studied sites, wheat showed highly significant response as depicted by sharply rising yield curves, indicating that N was the most yield limiting nutrient followed by S. Also, always increasing yield gains as observed in the graphs indicate a positive synergy that is occurring between the applied three nutrients, particularly N and S. But, as TSP contains mainly P and Ca as nutrients and others like S as impurity, its effect particularly on alkaline soils of ES zone may need further investigations.