4.1. Effect of Fertilizer Treatments on Maize Yield
The overall maize yield increase after additions of selected secondary macro- and micronutrients (SMNs) above the standard N, P and K fertilizer confirms that the closure of maize yield gaps in poorly responsive soils requires more than the 3 primary nutrients [
19]. On average, the smallest yield increase above the control was obtained after using the standard N, P and K fertilizer compared with using the NPK amendments for both regions. In Bungoma-Southwest, the largest yield increase of 2.3 t ha
−1 was obtained after using NPK3. This implies that, beyond the supply of N, P and K, sites in that region require both Ca, Zn and Cu at optimal rates. It is observed that sites in Bungoma-Southwest had strongly acidic soils (mean pH = 5.2) compared to those in Busia-North (mean pH = 5.5) (
Table 2) using the criterion documented by Kanyanjua, et al. [
41]. Neutralization of acidity by CO
32−, OH
−, and HCO
3− derived from limestone in Mavuno fertilizer (NPK3) may have resulted in more nutrient availability for maize crop uptake in such soils [
42,
43]. For sites in Busia-North, application of FYM with N and P inorganic supplements performs better than the other amendments. This is an indication that, combining N, P, K and Ca with small rates of Zn and Cu is adequate to restore the productivity of poorly responsive soils in that region. Nearly of all sites in Busia-North had more than 75% sand and, therefore, the NPK2 deems important for nutrient and water retention.
The conspicuous seasonal effect on yield also confirms weather pattern as a major constraint to maize production beyond nutrient deficiencies for the rain-fed agriculture [
44]. In general, during the short rains, maize yields reduced by 28% below the 3.6 t ha
−1 obtained during the long rains. Although a comparison between the yields at region level showed no significant differences, yields obtained during the second cropping seasons differed. Sites from Bungoma-Southwest produced larger yields after addressing the SMN deficiencies compared to Busia-North. It was observed that Bungoma-Southwest received a substantial amount of rainfall during the SR (cumulative rainfall of 482 mm between September and November) compared to the average of precipitation recorded over several years (
Figure 2). The rainfall increase was largest in the month of November which coincides with grain filling [
45] and consequently may have contributed to the yield difference between the sites
4.2. Maize Agronomic Response Clusters
Beyond the yield averages, the clusters reveal diverse maize response patterns across the study sites. Sites in Cluster 1 showed a response to all three NPK amendments. Soils from sites belonging to this cluster were not only deficient in primary nutrients, P (<10 mg kg
−1) and N (<0.2%) (
Table 6) but also in SMNs (Ca, Zn and Cu) and hence the response to all NPK amendments. Therefore, these results indicate the need of supplying together all nutrients that limit maize production in poorly responsive soils. In agreement with earlier reports [
23,
24,
46], on the contribution of SMNs in closing the yield gaps in SSA, it is also obvious in these soils. In addition, supplying small amounts of micronutrients, like Zn and Cu through FYM may be adequate for maize production in soils such as those found in this cluster. Negligible differences in yield improvement were observed from the application of FYM compared to the relatively high micronutrient rates supplied through the inorganic amendments. Soils for this cluster mainly comprised of Luvisols, Lixisols, Cambisols and Planosols. Those soils are commonly known for their relatively large base saturation and hence, can hold a larger amount of nutrients [
28]. Alongside the supply of nutrients, FYM may have further increased the nutrient storage capacity of these soils considering their average low organic carbon content and a large sand content of 71% [
40,
47]. A similar observation by Zingore, et al. [
48] also highlights the need of FYM for restoring productivity in nutrient depleted sandy soils.
Sites in Cluster 2 showed a selective response to the NPK3 amendment. Sites in this cluster also had low P (6.9 mg P kg
−1) - while slightly higher- compared to those in Cluster 1 (6.5 mg P kg
−1). Soil N was also low in sites belonging to Cluster 2 (
Table 6). Although the restricted response to NPK3 indicates a larger demand of micronutrients compared to sites belonging to other clusters, it also demonstrates the need of Ca application in such soils. In addition, the small S, Mg and Cu contents measured in maize ear leaves from sites in this cluster may account for the relatively low yield improvement (
Table 5). With reference to the S sufficiency ranges between 0.16–0.2% given by Reuters and Robinson [
49], S was clearly deficient for optimal maize production in sites belonging to this cluster. With reference to
Table 5, N and S play a significant role in allocating sites in this cluster, Cluster 2. The content of N requires being reduced while S requires being increased for those sites to move to Cluster 1. This implies an imbalance between N and S. According to FAO [
50], and in line with first principles, some nutrient deficiencies may be aggravated by application of another. Other studies in SSA also indicate that soils become deficient in nutrients like S once the macronutrient status has been optimized [
51]. The severity of S deficiency is usually aggregated by high rates of N application ([
52]), probably as those applied in our study.
In recent past, crop deficiencies of S have been reported in cropping systems that have reduced anthropogenic S input and failure to replenishment S through fertilizer input to compensate exportation [
53]. In line with this, continuous application of Sulphur-free fertilizers may also induce S deficiency [
54] such as the case in this study. All the nutrient sources were S-free except for the negligible content contained in FYM (NPK2) and Mavuno fertilizer, used in the NPK3 treatment [
55]. The consequence of S-deficient conditions is an inefficient utilization of N, P and K fertilizers and the resulting poor profitability [
56].
Similar to S, Mg was not addressed in this study since it had not been diagnosed as a major problem limiting maize production in poorly responsive soils [
27]. However, based on the sufficiency ranges between 0.21 and 0.5% given by Reuters and Robinson [
49], the measured average content of 0.11% for sites in Cluster 2, indicates deficiency. According to Gransee and Führs [
57], Mg deficiency principally occurs due to an absolute small content in the soil or due to cation competition. Using the criterion given by Okalebo, Gathua and Woomer [
30], soils in this cluster had moderate exchangeable Mg content on average of 0.74 cmol
c kg
−1 and hence not limiting. Application of K may have therefore accentuated Mg deficiency through cation competition [
58,
59]. In addition, low Mg content in ear leaves from sites in Cluster 2 confirms such possibility.
Likewise, while we did address Cu deficiencies in this study, ear leaf contents of this nutrient were still relatively small in samples from sites belonging to Cluster 2. On average, the Cu content for the latter was 1 mg kg
−1 below the 9.19 mg kg
−1 obtained from sites in Cluster 1 (
Table 5). However, those Cu contents for Cluster 2 may not be regarded as deficient as such since their values are still within the sufficiency ranges between 6 and 20 mg kg
−1 specified by Reuters and Robinson [
49]. Nevertheless, ion competition may still explain a scenario of deficiency. Both Cu and Zn are bivalent cations known to compete for adsorption, i.e., Zn may have inhibited Cu adsorption at the root surface [
60]. This can be derived from the slightly larger Zn contents for sites in Cluster 2 (16.22 mg kg
−1) compared to 15.9 mg kg
−1 measured for those in Cluster 1 (
Table 5).
Poor responses to SMNs interventions was observed for sites in Cluster 3. This implies that application of NPK amendments had an insignificant effect on yield increase above control for Cluster 3. The average maize yield of 3 t ha
−1 obtained from control plots is indicative of relatively fertile sites in this cluster and hence may be considered as ‘fertile poor responsive cluster’. The soils in this cluster comprised Eutric Cambisols and Gleyic Arenosols with adequate levels of available P at an average of 17 mg.kg
−1 [
30,
61]. Nevertheless, ear leaf content of several micronutrients (S, Mg, B and Cu) for this cluster were small compared to those measured in Cluster 1. Out of the four micronutrients, only Cu was added in the NPK amendments. Although the correlation between soil available P and ear leaf Cu content indicate a synergistic relationship (
Table 7), Cu content still seems inadequate to result in a significant maize yield increase. In addition, the emerging S and Mg deficiencies for the same cluster may have also occurred due to similar conditions as those explained for Cluster 2. Furthermore, the marginal B deficiency at 4.5 mg kg
−1 in maize ear leaves [
62] for Cluster 3 significantly relates to the negligible yield increase above the control. As shown from multinomial logit analysis, N and Ca ear leaf content has an antagonistic effect on B content for the same cluster. Application of N may, therefore, have offset the N: B ratio in soil resulting in low B uptake [
63]. Boron is also known to have a close relationship with Ca. Application of Ca to sites belonging to Cluster 3 may have reduced the availability of B resulting in its low uptake [
64,
65]. A dilution effect is also another possibility of the observed small B ear leaf concentration for Cluster 3. This mostly occurs when large Ca concentration in plant tissue increases B demand due to close similarity in function [
66]. In addition, Ca may also influence uptake of B indirectly. Application of Ca in soils for sites belonging the poor responsive Cluster 3 may have increased the uptake of P as observed in
Table 5. Both B and Pare anions that have an antagonistic effect. Increase in P uptake reduces B uptake [
67] and hence the small content measured.
4.3. Economic Benefit from NPK Amendments for Maize Grain Production
The Economic benefit of fertilizer use is affected by fertilizer cost, grain prices and ultimately how maize responds to fertilizer application [
22]. Investing on any of the NPK amendments at an average
$0.5 extra above the
$1.6 (
Table 3) already used to purchase the standard N. P and K fertilizer results in 3 times net profits. Such profits would be satisfactory incentives for investing in fertilizer use in SSA [
68,
69]. Moreover, it is important to identify the most profitable and suitable fertilizer intervention that fits a local context. In this case, results indicate that NPK3 (Mavuno based amendment) was more profitable in Bungoma-Southwest compared to the other amendments while the FYM based amendment was the most profitable in Busia-North. Further, delineating the type of maize responses clearly separates sites where NPK amendments can be recommended from those that still require further attention. In agreement with Kihara, et al. [
70] substantiating the highly variable profitability of fertilizer use helps farmers to make well-informed decisions on fertilizer use.
Application of the CART tool reveals the underlying soil characteristics that would predict the profitability of fertilizer use. The tool, therefore, provides a simple method of determining which would the most profitable fertilizer interventions under specific soil conditions. Use of the standard N, P and K fertilizer can be profitable only to farmers whose sites have less than 5.04 mg P kg
−1 irrespective of the response clusters. Only 28% of the total sites are in this category; confirming the diagnosis of poorly responsive soils in an earlier study [
27]. Obviously, the use of NPK amendments remains a risky intervention for farmers with sites that belong to the poorly responsive, Cluster 3. Therefore, recommending the NPK amendments would also not be appropriate to farmers whose sites belong that cluster. However, determination of judicious and balanced nutrient combinations for maintaining soil productivity in such sites is indispensable [
71,
72]. For example, applying lower N and Ca rates may not only reduce the fertilizer cost but would maintain the desired nutrient balance ratio in maize ear leaf tissue. Application of NPK amendments can be beneficial to farmers whose sites belong to clusters 1 and 2, respectively. Furthermore, the smaller the available P (< 4.72 mg kg
−1) in such soils, the more financial benefits may be realized. However, if the soils have more than 4.72 mg P kg
−1 of soil, supplying of both macro and micronutrients at optimum levels is critical for sites in Cluster 2. For maximum benefit of the NPK amendments, sites should have less than 2.93 mg P kg
−1 of soil. In such case, not only P would be limiting maize production but also the secondary and micronutrients.