Understanding nutrient imbalances in maize (Zea maysL.) using the diagnosis and recommendation integrated system (DRIS) approach in the Maize belt of Nigeria

Low nutrient use efficiency in maize as a result of imbalanced nutrition has been reported to drastically reduce yield. We implemented a nutrient omission experiment to assess the effect of nutrient application on maize yield and nutritional balance. Maize ear leaves were analyzed for nutrients, to identify nutrient balance status using the Diagnostic and Recommendation Integrated System (DRIS) approach. Results indicated that omission of N or P resulted in highly imbalanced DRIS indices respectively, and significantly lower grain yield. A strong inverse relationship between K ear leaf content with DRIS index suggests that K application negatively increases K imbalance in many situations. Imbalances of Mg, Ca and Cu were more associated with higher yielding treatments. A Which-Won-Where result show that nutrient imbalances in the diagnosis were systematically frequent when N was omitted. All the diagnosed nutrients were imbalanced even under the highest yielding NPKZn treatment; indicating further opportunity for yield increase with more balanced nutrition. Balanced nutrition of maize in the maize belt of Nigeria should target application of varying rates of N, P, K, Mg, S and Zn, depending on the soil conditions. But, because of complexities of nutrient interactions during uptake, it is hardly possible to realize a balanced nutrition. However, differentiating the application of antagonistic nutrients into foliar or soil-based methods is recommended for a more balanced maize nutrition.

Experimental design and treatment structure. A nutrient omission experiment was established in thirty farms selected across the study area, where each site was considered as a replicate. Compared with conventional nutrient omission experimental design where nutrient omission is based on macronutrients; N, P and K (Kihara et al. 49 and Shehu et al. 5 ) so that nutrient omission combinations results in NPK (no nutrient is omitted/all nutrients applied), NP (K omitted), NK (P omitted) and PK (N omitted) and a Control treatment (all nutrients omitted/no nutrient applied), we used a special nutrient omission approach which had not only the N, P and K omission treatments, but also NPK+ treatments in which micronutrients S, B and Zn were subsequently added to NPK, resulting in NPKS, NPKB, and NPKZn treatments accommodated in the omission design. Moreover, we added an NPKSZnB treatment which contained all the nutrients in the design. Together this can also be considered a nutrient omission trial both for macro and micronutrients. The addition of NPKSZnB in the designed allows for the evaluation of the degree of micronutrients limitation, besides evaluating the limitation in macronutrients. The plots were laid out in a randomized complete block design; with 3 blocks each consisting 3 plots. Each plot consisted of 6 rows (spaced at 0.75 m apart) of 5 m length to give a gross plot area of 22.5 m 2 . The plots were separated by an unplanted row or 1 m when row spacing was not applicable. The nutrients N, P and K application rates were 140 kg N ha −1 , 50 kg P 2 O 5 ha −1 and 50 kg K 2 O ha −1 respectively. Nitrogen (N) was applied as urea (manufactured by Notore Chemical Industries Plc, Nigeria) in three equal splits, i.e., at planting (basal), at 21 and 42 days after emergence (DAE), while full doses of the other nutrients were applied during planting. P was applied from Triple super phosphate (TSP-manufactured by Elixir Garden Supplies), while K was applied using muriate of potash (MOP produced by Gold Prime Fertilizer Company, Nigeria). Elemental S (locally sourced as pyrite) and Zn (using ZnSO 4 manufactured by Park scientific limited, UK) were applied at the rate 10 kg ha −1 while B was applied at 5 kg ha −1 using Borax (U.S. Borax Inc. USA).
Maize variety, sowing and harvesting. SAMMAZ  www.nature.com/scientificreports/ into stems and cobs, and dried. The dried cobs were shelled and the grains weighed for yield determination. The grain yield was expressed in kg ha −1 after weighing the shelled grains and adjusting the moisture to 15% using grain moisture tester as reported by Kamara et al. 2 .
Soil and Ear leaf sampling and analysis. Three soil samples from each site at 0-20 cm sampling depth were collected using soil auger prior to trial establishment in a 'V-shape sampling approach' . The samples were then placed in a basin, thoroughly mixed and a composite sample obtained. The composite samples were air dried before chemical analyses at the analytical service laboratory of the International Institute of Tropical Agriculture (IITA) Ibadan, Nigeria. Soil pH water in water (S/W ratio of 1:1) was measured using a glass electrode pH meter and the particle size distribution following hydrometer method. Total soil organic carbon (SOC tot) was measured using a modified Walkley-Black chromic acid wet chemical oxidation and spectrophotometric method. Total nitrogen (N tot) was determined using a micro-Kjeldahl digestion method 31 . Available Phosphorus (P avail.) and Sulphur (S avail.), exchangeable cations (K, Ca, Mg and Na) and micronutrients (Zn, Fe, Mn and B) were determined using Mehlich-3 extraction procedure 32

DRIS analysis.
For establishing the DRIS norms, the data was divided into two sub-populations i.e., a low (non-reference) and high yielding (reference) sub-populations. This was done by first sorting the plots according to yield in decreasing order and then portioning the data into the two sub-populations. The high yielding population comprised plots having grain yield higher than the Mean + 0.5* standard deviation; the mean and the standard deviation were calculated from the entire dataset. Thus, plots with yields ≥ 3766 kg ha −1 were considered high yielding, and they constituted 31% of the entire dataset. DRIS norms and coefficients of variations (CVs) were derived according to the procedure by Walworth and Sumner 25 . Mean value for each nutrient pair (example, A/B), their corresponding CV, and variance (σ 2 ) were then calculated separately for the two subpopulations. The mean value of each dual nutrient ratio in the high-yielding population were used as DRIS norms 25 . To select the dual nutrient ratio order (A/B or B/A) for calculating the DRIS indices (Eqs. 1 and 2, or 3 and 4), the variance of the low yielding sub-population was divided by that of the high yielding sub-population. It is theoretically expected that the data of the low yielding population would be more unbalanced and therefore should have larger variance than the high yielding one. The dual nutrient ratio with higher variance ratio value were therefore selected and used in the diagnosis. The DRIS indices were calculated based on ratios of each nutrient relative to all other nutrients using the equations below by Walworth and Sumner 35 . The DRIS index is the mean deviation of the dual nutrient ratios of a given nutrient from their respective norm values. The nutrients were ordered based on their DRIS index value, to determine the order of the degree of limitation to produce the nutrient limitation diagnosis. A DRIS index value for given nutrient close to zero ("0") indicates nutritional balance for that given nutrient relative to other nutrients in the diagnosis. A more negative index value for a given nutrient indicates lower amount or shortage relative to other nutrients. Alternatively, higher positive index value of a nutrient indicates excess presence of that nutrient relative to others 36 .
If we consider hypothetical nutrients A, B through N, then: www.nature.com/scientificreports/ where a/b is the DRIS norm for the ratio of nutrients A and B, and CV is the coefficient of variation associated with that norm expressed as percentage. A/B denotes the ratio of average concentration of the ten ear leaves collected per plot for nutrients A and B, n is the number of nutrients considered, and f (A/B) is a function of nutrients A and B ratio. The 1000 multiplier in Eqs. 3 and 4 comprises of a factor 10 to give the resultant indices a convenient magnitude and a factor 100 to express the CV as fraction rather than as percentage.

Statistical analyses.
Descriptive statistics was carried out to study the distribution of the soil, yield and ear leaf data using mean, range and coefficient of variation (CV). Relationship between respective nutrient ear leaf content and DRIS index was studied using a regression analysis, with nutrient ear leaf content as the independent variable and the DRIS index as the dependent variable. Simple scatter plot of the DRIS indices was plotted to study the relationships among the nutrient DRIS indices. Comparison GGE biplot was drawn with sectors to demarcate most responsible nutrient treatment which influenced nutrient DRIS indices based on mean and stability. All statistical analyses were done in GenStat 17th edition statistical package (VSN International).

Results
Physico-chemical soil properties at the experimental sites. Soil properties across the experimental sites showed wide variation, except for pH which has a low coefficient of variation (CV) of 7.90% (Table 1). Average pH value (6.63) indicated that the soils can be generally regarded as slightly acidic. Total soil organic carbon (SOC tot ) was very low, and this condition seems to be moderately consistent across the sites. Average total nitro- Maize grain yield. Maize grain yield ranged from 101.00 to 7450 kg ha −1 across the nutrient treatments (Tables 2 and 3). Except under NPKB plots, average grain yield of the NPK and NPK + treatment plots were above 3,000 kg ha −1 . The NPKZn treatment increased grain yield by 28% compared to NPK. Omission of either N or K (in PK and NP treatment respectively) both reduced yield by 59.00%, and that of P (NK treatment) reduced yield by 56.00% compared to that of the NPK treatment. On average, yield for the Control plot was below 1500 kg ha −1 . The highest coefficient of variation (66.00% and 65.00%) of yield under Control and NP, respectively indicated that the experimental fields were highly variable in terms of fertility status.

Ear leaf nutrient concentration. Ranges of nutrient ear leaf concentrations across the treatments is
shown in Tables 2 and 3. All treatments with N application resulted in higher N ear leaf concentration compared to the Control and PK treatments. Phosphorus (P) concentration ranged from 0.04% in Control treatment to 0.39% in NPK and NPKZn treatments. Higher average P ear leaf concentrations were observed for the NP, PK and all NPK treatments, and were significantly different from that of the Control (0.16%) and the NK (0.20%). For K, there was no significant differences between the various treatments, though the NK and NPK + treatments had slightly higher ear leaf K concentrations. The omission of K did not reflect in lower average ear leaf concentration. The highest average K ear leaf concentration was observed for the NK treatment. There was no indication that K application in itself led to higher K concentrations in the ear leaf. Average ear leaf concentration of Ca was below 1.23% across the treatments. However, Ca concentration was higher in NPK and NPK + treatments. Magnesium (Mg) concentration ranged from 0.01 to 1.46%. There was little difference in the average Mg and Ca concentrations for the various treatments, only that the Control had a lower concentration for the elements. Sulphur concentration ranged from 0.01 to 1.29% across all treatments. With the exception of NPKSZnB treatments (0.05%), the S ear leaf concentration for all the NPK related treatments was slightly elevated (ranging from 0.71 to 0.84% on average) compared to the Control, NP, NK and PK treatments, which seems to suggest some relationship to the yield level. There was no clear link between the micronutrients in the ear leaf to any of the treatments. Highest concentration of Cu was 11.25 mg kg −1 for the Control treatment and the lowest concentration of 2.24 mg kg −1 was found

DRIS indices.
The DRIS ratios were selected based on higher variance ratio between the low and high yielding sub-populations. The variance ratios of all the selected dual ratios of the low against the high yielding subpopulations were ≥ 1, which indicate relative higher variance of the low yielding sub-population. The variance ratios of N/K (153.72) and B/Zn (117.97) (Table 4) were the highest among the ratios. Percentage of plots with negative DRIS indices for the nutrient treatments (Table 5) indicates that PK, NPKS and NPKZn treatments have highest percentage of plots with negative DRIS nutrient indices. Among the nutrients, B had the lowest percentage of plots with negative index values across the nutrient treatments, while Fe had the highest number of negative index plots. Among the nutrient omission treatments, the PK treatment showed the highest frequency of plots with negative N index which reflects the impact of the N limitation on the DRIS index value. The NK, NP and NPK treatments showed lower percentage (33.30 and 40.00% respectively) of plots with negative N DRIS index indicating that N was imbalanced due to higher concentrations of other nutrients. The percentage of plots with negative N DRIS index for the NPK and NPK + treatments ranged from 40.00 to 64.30% apart from the NPKS treatment which had 84.60% of plots with a negative N DRIS index value. The slightly higher percentage for the NPKZn treatment indicates a possible dilution of the N concentration on the ear leaf because of the higher yield.
The percentage of negative P DRIS index value plots was lowest for the 'Control' treatment (viz. 58.30%) and for the other treatments the percentage did not differ much, with the percentage ranging between 64.30 to 84.60%, indicating that for all treatments the relative P concentration is lower than for the reference population. There was no marked decrease in percentage of the plots with negative P DRIS Index value for the NP and PK treatments or a marked increase in percentage for the NK treatment (P-omission treatment).
Omission of K (NP treatment) resulted in a reduced percentage (33.3%) of negative K index plots, contrary to what would be expected and the NK treatment showed an unexpected high percentage of plots with negative K DRIS index values. Otherwise, the percentage of negative plots varied between 46.75 and 76.90% indicating that the relative K concentration was generally lower than for the norms where K was involved. The NPKS increased number of plots with negative Fe index by 19.00% and that of B by 13.00% compared to NPK. Ca index however, becomes increasing negative in plots where Zn was applied to or N was omitted from NPK. Both the PK and NPKZn treatments resulted to relatively higher percentages of plots with negative DRIS index scores for most of the nutrients. For the NPKZn, this was also true for most nutrients except for the N and P DRIS indices and to lesser extent for the S DRIS index score. The generally lower percentage of plots with negative B DRIS index values, across the various treatments, indicates that the relative B ear leaf concentrations of the low sub-populations was not much different, which probably indicates that the B concentration was low across the board. On the other hand, the percentage of negative plots with negative Fe DRIS index values was relatively high, irrespective of the treatment, indicating that the reference population of plots with high yields have relatively lower Fe concentration in the ear leaves. The same applies to Mg, where we find high percentages of plots with negative DRIS scores across the treatments, indicating that the reference population has a relatively low Mg ear leaf concentration, indicating a possible nutrient limitation. Table 6 shows the ranking of the nutrient limitations based on the DRIS index value for each treatment. For the interpretation one has to take account of the treatment and the nutrients. N ranked high (2nd) in the order of nutrient limitations for the PK treatment, while it ranked lower for other treatments where N was applied, indicating N was highly limiting. For P and K, the pattern was less clear and also not ranking highly in terms of nutrient limitations for the respective nutrient omission treatments. Sulphur (S) ranked high in the order of nutrient limitations for the NK, NP and PK treatments as well as for the NPK and NPKSZnB treatments. It ranked low for the NPKS and the NPKZn. This suggest that S was clearly a limiting nutrient, with the low ranking for the NPKZn explained by the possible positive effect of Zn application on the availability of S. Zinc (Zn) ranked high in the order of nutrient limitations for the 'Control' , NK and NP treatments, but did not seem to be specifically www.nature.com/scientificreports/ prominent for the other high yielding treatments. Boron ranked lowest based on its DRIS index value for seven of the nine treatments and ranked highest for the NPKZn treatment. Note that the NPKZn treatment generated the highest yields and the corresponding plots will therefore constitute a large part of the reference population. The negative interaction with B thus explains the generally low percentage of plots with negative B DRIS index scores and consequently low ranking in the order nutrient limitation for the remaining treatments other than NPKZn. Therefore, the low ranking in order of nutrient limitation in this case does not necessarily signify that B is not a limiting nutrient.
With regards to the nutrients that are not included in any of the treatments, Mg ranked high in the order of nutrient limitations for all the treatments (often ranking first, second or third). This is also consistent with the relatively high percentage of plots with negative Mg DRIS index values. Therefore, among this group Mg is considered the most important yield limiting nutrient. Secondly, Fe ranked relatively high, ranking second to sixth depending on the treatment. Also, in this case it is consistent with the generally high score of plots with negative Fe DRIS index values across the treatments. No particular evidence of Mn limitation was observed in any of the treatments. Noteworthy is the high ranking in order of nutrient limitation for the NPK (3 rd ) and the NPKZn treatment (2 nd ), indicating possible interaction with Zn application. A similar observation was made for www.nature.com/scientificreports/ Ca and Cu, in that there was no particular evidence that the nutrients might be limiting, but that for particular treatments the high ranking is noteworthy. In this case the NPKS, NPKB and the NPKSZnB treatments show high ranking of Ca and Cu (Table 6) suggesting an interaction with S and B on the availability of these nutrients.   www.nature.com/scientificreports/ GGE biplot analysis. Figure 2 shows relationship among the nutrient DRIS indices on a biplot. PC1 explained 36.32% of the total variation in the data. The PC1 was strongly influenced by the positive correlations with the DRIS indices Mn, N, Fe, P and B. Additional 20.32% of the variance was explained by PC2, and was positively correlated with index of Zn with negative correspondence with Cu index. Indices of P, B, Mn, N, B, K, and S were positively correlated with each other; indicating a synergistic relationship. Indices of Zn and Mg were positively correlated with each other and negatively with Cu and Ca, indicating a possible antagonistic association between them. The 'Which-Won-Where' view of the GGE biplot (Fig. 3) show the degree of influence of nutrient treatments on the nutrients DRIS indices. The general rule in GGE is that nutrient indices that share the same sector with a particular nutrient treatment are the most associated with that treatment. The indices of Zn and Mg formed a cluster and share the same sector with Control and NK treatments. The largest cluster was formed by the indices of K, S, P, Mn, Fe, B and N, and found in the sector of PK treatment. This sector is bordered closely by Control, indicating that these two treatments have similar DRIS index values for most of the nutrients that are within that cluster. In the diagnosis, Cu and Ca indices became more consistently important yield limiting under NPKZn and NPKSZnB. The DRIS nutrient indices associated with NPK, NPKS and NPKB did not show any consistent pattern of occurrence.

Discussion
Maize yield was significantly affected by the nutrient treatments. The results reflected the responsiveness of maize to the applied nutrients compared to when no nutrient was applied. As reported by Shehu et al. 5 , maize yield can reach > 5 tons per hectare in northern Nigeria when nutrient deficiencies are properly addressed. Application of NPKZn gave the highest yield. This indicates relative response to Zn in the study area. The result confirmed the findings of Garba et al. 40 , that Zn application increased maize yield in some parts of Nigeria. The low CV of yield under NPKZn treatment in this study also confirmed the stability of maize yield response to Zn in the study area. This may be due to the improvement of fertilization uptake caused by Zn application 41 . Yield decrease due to B application observed in this study has also been previously reported by Shehu et al. 5 . Results of the soil analysis also confirmed B deficiency in the soils in this study, however, since there is dearth information about B application rates, it is possible that the B application rate used in this study was higher than required, and might have resulted to B toxicity, or lower than required for significant B response. In this sense it can be argued that the high amount of soil Zn might have countered B expression as similarly reported by Rehman et al. 42 .
Occurrence of negative DRIS index of N was higher in the treatments where no nitrogen was applied (Control and PK) or when other nutrients were applied with NPK (especially NPKS and NPKZn). A significant increase in ear leaf concentration of N beyond the critical level of 2.60% when N was applied also indicated that maize responded to N treatments. The relationship between N DRIS index and N ear leaf concentration discovered in this study indicated a proportionate increase in N balance index at higher ear leaf concentrations. But since this relationship is weak, emphasis could not be made concerning the negativity of the N balance index in some plots where N was applied. In a similar study, Reis Júnior et al. 43 , Silveira et al. 44 , Nachtigall and Dechen 45 observed very low correlation coefficients for N and S between their ear leaf concentrations and respective DRIS balance indices, indicating that other important factors play role in the nutrient balance. The current study was on-farm that cut across wide range of cropping potentials. This could have possibly caused a large variation in the ear leaf content of nutrients due to different previous nutrient management which affected the N balance. In progressive diagnosis, Sumner 46 discovered that negative N index was not an indication of deficiency, but an imbalance caused by continuous application of P, S and K. Similarly, the high percentage of negative N index plots of NPKS and NPKZn treatments (compared to the NPK) might not directly be related to N deficiency, rather an imbalance as a result of S and Zn additions. This can be attributed to the very low variance between the dual nutrient ratios (N/S and Zn/N) of the sub-populations 47 . It can therefore be concluded that N diagnosis using these treatments is likely ineffective in this study. The K ear leaf contents for NK and NPKZn were within the sufficiency range (1.20-1.70%) for the region reported by Reuter et al. 48 , this thus indicates some level of K response. However, K application under PK and NPK, showed a significantly lower K leaf concentration below the critical limit. Therefore, it can be concluded that K response is likely a site-specific scenario. The strong inverse relationship between K ear leaf content with DRIS index indicates that K application negatively increase K imbalance in many situations. Further, the table for order of nutrient limitation indicated that K was limiting in both scenarios of omission and application. As indicated by Kihara et al. 49 response to K may be very patchy. Though the lower yield of NP observed indicates the significance of K, the large variation in the results (CV is 65%) further confirmed variable response to K in the area. Shehu et al. 5 and Nziguheba et al. 10 both indicated that K is not a limiting nutrient for maize production in some soils within the study area. Inconsistent response to K could be attributed to the large deposit of K-feldspars as the dominant mineral 28 in rocks of the basement complex and dust deposition of K during the harmattan 10 which resulted in inherent high K concentrations in the soil, but these may have been depleted depending on the type of land use and land use history and caused variable response.
The ear leaf concentration of P was below the critical limit (0.27%) established by Reuter et al. 48 , apart from those of NPK and NPKS treatments (0.28% and 0.27% respectively). Effect of P application did not directly reflect in the P balance index except for the NPKS where P was the second least important limiting nutrient. The higher ear leaf content and lower limitation order of P under the NPKS is not surprising as P and S interact in synergy to influence uptake of each other. Studies in Nigeria by Ogunsola and Adetunji 50 recommended S application for enhanced P uptake by maize. The high recurrence of negative DRIS P balance index and increased order of P limitation for NPK and NPKB treatments is more difficult to explain, especially as P uptake is reported to be positively influenced by B application 51  www.nature.com/scientificreports/ results, it is likely that rate of B applied in this study might have affected the P balance. The strong positive correlation between P and B indices in this study further confirms the effect of B on P. Also, most of selected dual nutrient norms that involved P for the high yielding sub-population (which were used for the diagnosis) were not much higher than their corresponding low yielding sup-population. This means that the DRIS index calculated for P involved some samples from high yielding but unbalance sub-population, and this must have affected the P DRIS index. Though the norms involving P in this study are in most cases higher than those obtained by others like Serra et al. 52 and Anabela et al. 53 in other continents, yet the ear leaf P contents were below the sufficiency limit, this therefore suggest that nutrient norms cannot be universally adopted.
Range of Ca concentration in the ear leaves was small as indicated by a low CV (< 25%) for most of the treatments. The average concentration of Ca was obviously above the critical concentration levels of 0.21% and 0.30% established by Reuter et al. 48 and Nziguheba et al. 10 respectively. Although, Ca was not a treatment in this study, some amount of Ca (~ 15%) is contained in TSP fertilizer that was used as source of P in this study, might be the reason for the high Ca ear leaf concentrations, especially where P was applied. Negative DRIS Ca index is also confirmed to be associated more with the NPK treatments as shown in the Which-Won-Where plot. The percentage fields with negative Ca index was also lowest in the NK confirming the earlier suggestion. In the WBE experiment, Nziguheba et al. 10 also reported negative Ca index, even though ear leaf concentration of Ca was above the critical limit, and they concluded that the negativity indicated Ca was imbalanced not deficient. Therefore, we conclude also that the negative Ca index observed in this study translates to imbalance. This is further confirmed by the high level of Ca content in the soils observed in this study. Adequate amount of Ca was reported by Agbenin 54 , in most soils in West African savanna except in cases of long-term continuous cultivation and without Ca application. Negative S index was more associated with PK treatment and less with NPKS and other treatments. Contrary to the conclusion made by Nziguheba et al. 10 that pronounced imbalances of Mg and Ca hide S effect, our results revealed that S is next most critical nutrient after deficiencies of N, P and K were addressed.
The average B ear leaf concentration for all the treatments was within the sufficiency range of 2-5 mg kg −1 established for maize by Kelling 55 . The B treatments resulted in relatively higher B DRIS index. However, it did not reflect in the B leaf concentration, which was relatively lower for the NPKB treatment. The results were similar to that of Aref 51 , where the Control and B treatments were not significantly different for B ear leaf concentration. The occurrence of negative index of B across the treatments was however lower compared to those of the other nutrients. Boron (B) was also shown to be least important limiting nutrient except under NPKZn. Though previous studies 56 have showed widespread deficiency of B in the soils within this study area, but considering the very narrow range (0.3-1 mg kg −1 ) between deficiency and excess levels for soil B, it could be assumed that the B application rate (5 kg ha −1 ) used in this study might have been high enough to have caused B toxicity, and subsequently, the low yield of the B treated plots. The Zn ear leaf concentrations were higher than those reported by Nziguheba et al. 10 in the same region. Our findings seem to suggest an effect of Zn application on the ear leaf content, with the NPKZn and the NPKSZnB treatments showing slightly elevated Zn ear leaf concentration among the NPK treatments. The Zn ear leaf concentration for the NK and NP treatments were relatively high (though not significantly different from the other treatments) which also seems to be reflected in the higher Zn DRIS index values (− 8.4 and − 3.1 respectively-not shown). The pronounced limitations by Ca and Cu seem to have masked the effect of Zn in many of the treatments. Other analyzed micronutrients (Cu, Mn and Fe) in this study were mostly above their critical concentration levels with all treatments except when N was not applied, or when B was applied.

Conclusions
The nutrient treatments used in this study significantly affected the ear leaf nutrient concentration, DRIS indices, and maize yield. Omission of any of the major macro nutrient (N, P and K) resulted in lower yield than when they were applied. Ear leaf concentrations of most nutrients correlated well with their corresponding DRIS indices. The DRIS method identified nutrient imbalance for maize production in the maize belt of Nigeria. The diagnosis indicated consistent imbalances of unused secondary macro and micronutrients (Mg, Cu and Ca) in most NPK treatments; revealing their relative importance in maize nutrient management in the area. The Which-Won-Where showed that the frequent occurrence of negative imbalances of nutrients associated with PK treatment is systematic rather than random. None of the diagnosed nutrients is within the balance range of − 1 to + 1 even under the highest yield treatment (NPKZn); indicating that yield can be further increased when nutrient imbalances are further addressed. DRIS diagnosis is complementary to soil analysis in diagnosis of nutrient limitations and appropriate method to making recommendations on balanced crop nutrition. Balanced nutrition of maize in the maize belt of Nigeria should target the application of varying rates of N, P, K, Mg, S and Zn, depending on the soil condition. Because of complexities of nutrient interactions during uptake, it is hardly possible to realize a balanced nutrition. However, differentiating the application of antagonistic nutrients (especially the divalent cations) into foliar or soil methods is recommended for a more balanced maize nutrition.