Plant density and nitrogen responses of maize hybrids in diverse agro-ecologies of West and Central Africa

: Maize ( Zea mays L.) production in West and Central Africa is constrained by drought, low soil-N and Striga infestation. Breeders in the region have developed and commercialized extra-early and early-maturing hybrids (E-EH and EH), which combine high yield potentials with tolerance/resistance to the three stresses. Hybrids of both maturity groups are new to the farmers; thus, the urgent need to recommend appropriate agronomic practices for these hybrids. We investigated the responses of four hybrids belonging to extra-early and early-maturity groups to plant density (PD) and nitrogen (N) application in five agroecologies. The EHs consistently out-yielded the E-EHs in all the five agroecologies. The hybrids showed no response to N-fertilizer application above 90 kg ha -1 . All interactions involving N had no significant effect on all traits except in few cases. The E-EHs and EHs had similar response to PD; their grain yield decreased as PD increased. Contrarily, flowering was delayed and expression of some other agronomic traits such as plant and ear aspects became poorer with increased PD. Optimal yield was obtained at approximately 90 kg N ha -1 and 66,666 plants ha -1 . Most of the measured traits indicated high repeatability estimates (i.e. ≥ 60) across the N levels, PDs and environments. Evidently, the hybrids were intolerant of high PD.

(IITA) and collaborators have developed and released extra-early and early-maturing maize hybrids that combine high yield potentials with tolerance/resistance to low soil-N and drought at flowering and grain filling periods, and are being adopted by the farmers in the sub-region. However, the responses of these hybrids to plant density and N application has not been investigated. It is desirable to investigate and document grain yield response to plant density and N application of these hybrids released for the different agro-ecologies in WCA. Such information can guide future breeding of new cultivars and cropping technique innovation. This study was conducted to: (i) investigate the response of grain yield and other agronomic traits to plant density and N rates of recently released four extra-early and early-maturing drought tolerant (DT) maize hybrids and the performance of each maturity group in different agro-ecologies of Nigeria, and (ii) partition the total variation in grain yield to its various components.

Experimental sites
This study was conducted during the growing season of 2015 at five locations, one location in each

Germplasm and experimental design
Four DT maize hybrids (belonging to extra-early and early maturity groups) recently released in Nigeria, Mali, and Ghana were evaluated. For each maturity group, one single-cross (SC) and one top-cross (TC) hybrids were evaluated in the study. The hybrids are tolerance to low soil-N, drought and resistant to Striga with high yield potentials [20]. The detailed descriptions of the hybrids are presented in Supplementary Table 2. Each experiment in each location was grown in a randomized complete block with a split-split-plot arrangement and three replications. The main plots were the N fertilizer rates (90, 120 and 150 kg N ha -1 ), plant densities (66,666; 88,888 and 133,333 plants ha -1 ) were subplots, and the four hybrids were sub-subplots. The inter-row spacing of 0.75 m and three intrarow spacings of 0.4 m, 0.3 m and 0.2 m were used to obtain the three plant density levels. Each subsubplot comprised four rows, 5 m long each. Three seeds were planted per hill and thinned to two plants/stand two weeks after emergence. Urea was the source of N.

Data collection
Observations were made on the two central rows within each sub-subplot. Data obtained included anthesis (ANTH) and silking (DYSLK) which, respectively were the number of days from planting to the date when 50% of the plants in a sub-subplot had shed pollen and emerged silks. Anthesis-silking interval (ASI) was computed as the difference between DYSLK and ANTH. Plant height (PLHT) and ear height (EHT) were measured as the distance from the base of the plant to the height of the firsttassel branch and to the node bearing the upper ear, respectively. Root lodging (RL) was determined was based on freedom from disease and insect damage, ear size, uniformity of ears, and grain filling and was scored on a scale of 1 to 9, where 1 = clean, uniform, large, and well-filled ears and 9 = ears with undesirable features. Husk cover (HC) was scored on a scale of 1 to 5, where 1 = husks tightly arranged and extended beyond the ear tip and 5 = open tip cover (ear tips clearly exposed). Field weight was recorded as the weight in kg of all de-husked ears (cobs) in the sub-subplot.
Representative cobs were selected and grains removed from their cobs. Moisture meter was used to measure the moisture of the grains. Grain yield (GYLD) in kg ha -1 was calculated based on 80% shelling percentage and adjusted to 15% moisture content as follows: Grain yield (kg ha -1 ) = Field weight × (100 -actual grain moisture %) /85 × {10000/plot area (m²)} × 0.80.

Statistical analysis
Analysis of variance (ANOVA), combined across trial environments was performed on plot means for the individual traits with PROC GLM in SAS using a RANDOM statement with the TEST option [21]. In the combined ANOVA, location and replication nested within location was considered as random effect for each trait, while N, plant density, and genotype, and their interactions were considered as fixed effects and interactions involving locations were considered as independent effect. Comparison between and within maturity group was achieved by partitioning the genotype sum of squares into orthogonal contrasts. Linear regression was fitted to quantify the grain yield and other traits responses to plant density using Microsoft Excel package. Each trait was considered as dependent variable and plant density as the independent variable. The proportion of total variation in grain yield accounted for by the different sources of variation in the combined ANOVA was manually computed by dividing the individual sum of square of each source of variation by the total sum of square, estimated in percentage. The estimates of broad-sense heritability (H 2 ) for grain yield were computed for each environment. All the environments included in the present study revealed an H 2 value of ≥ 0.30 (Table S1). The H 2 of grain yield was estimated as follows: where σge 2 is the variance attributable to genotype x environment effects, and e is the number of environments; σg 2 , σ 2 , and r as defined above

Field performance of extra-early and early hybrids and repeatability of traits.
The combined ANOVA showed highly significant genotype (G), environment (E) and G x E interaction mean squares for grain yield and all other measured traits (3a and 3b). The contrast analysis for the two maturity groups (extra-early vs early maturing hybrids) showed highly significant mean squares for all the traits. (Tables 3a and 3b). The between-group comparison (extraearly vs early maturing hybrids) for grain yield accounted for 56% of the variation among genotypes.
The early maturing hybrids consistently out-yielded the extra-early maturing hybrids in all the agroclimatic zones with an average of 428 kg ha -1 (16%) across environments ( x TZEI 7) in all the agro-climatic zones (Table 4). Also, root and stalk lodging were higher for the extra-early than for the early maturing hybrids, although; both lodgings were more pronounced at the forest than the savanna locations (Table 4). In each maturity group, root and stalk lodgings were higher for TC than for SC hybrids (Table 4). It is striking to note that both maturity groups received good scores (in the range of 3 and 5) for PASP in all the agro-climatic zones except for extra-early hybrids that received a poor score (6) in the MRF zone. For EASP, both maturity groups received poor scores (6) at the two forest locations and good scores (4 and 5) at the savanna locations ( Table   4).
The repeatability estimate of grain yield was 0.83. The repeatability estimates of other agronomic traits varied from 0.40 for EASP to 0.90 for DYSK ( Table 4). Most of the measured traits of the hybrids indicated high repeatability estimates (i.e. ≥ 60) across the three N levels, three PDs and five environments (Table 4).      Figure 1). Extraearly and early hybrids had similar linear trends for PD, although, the regression parameters were higher for the early than extra-early hybrids ( Figure 1). Rates of decrease in grain yield related to increased PD were roughly the same for all hybrids except the SC extra-early hybrid, TZEEI 29 x TZEEI 21, which had a lower rate relative to others ( Table 6). The R 2 value (0.0616) for the linear response of TZEEI 29 x TZEEI 21 to PD was far lower than those obtained from the linear responses of other hybrids ( Table 6). The response of EPP to PD was similar to that of grain yield; that is, it decreased with increased PD (Table 5). On average across PD, the number of EPP produced in the savanna locations were about 9% higher than those produced in the two forest locations. The trait (EPP) exhibited a negative linear response to PD (Table 6). In contrast, the mean values of PASP, HC, RL, and SL increased significantly with increased density, although the mean values were much lower in the savanna than forest locations ( Table 5). The respective linear trends in the PD response (averaged across the four hybrids, three N rates, and five locations) for PASP, EASP, HC, ASI, RL, and SL were all positive ( Table 6). The coefficients of determination, R 2 of the traits ranged from 89.29% for PASP and EASP to 100% for RL and SL (Table 6), indicating high reliability of the linear regression models for the traits. Interestingly, EASP and PASP had the same R 2 value (89.29%) and the rate of increase in score value (2E-05) associated with increased PD, but the intercepts were quite different about 3.8 and 2.8 for EASP and PASP, respectively (Table 6).

Response of grain yield and agronomic traits to N rates
The combined ANOVA showed no significant N mean squares for grain yield and all other measured traits (Tables 3a and 3b). The effect of the three N rates for grain yield across the three plant densities and five environments was the same for all the genotypes (Tables 3a and 3b). Although, the highest grain (2734.34 kg ha -1 ) across the four genotypes, three plant densities and five environments was obtained at 120 kg N ha -1 , but was not different statistically with grain yield (2681.43 kg ha -1 ) obtained at 90 kg N ha -1 . Thus, the hybrids showed no response to N-fertilizer application rates above 90 kg N ha -1 . N x PD interaction, as well as all other interactions involving N, had no significant effect on   all the studied traits except in a few cases such as SL (Tables 3a and 3b).

Partitioning of the total variation in grain yield into its various components
The E, G, and G x E interaction mean squares in that order, were the most important contributors determining grain yield. Environment, the largest contributor, accounted for 23.5%. The variance accounted for by the G and G x E interaction sum of squares were 19.4% and 6.4 %, respectively ( Table   7). Partitioning of the variance contribution of the genotype sum of square revealed that orthogonal contrast between maturity groups (extra-early vs early hybrids) accounted for 55.5 % and within each maturity group accounted for 26.7 % (SC vs TC extra-early hybrids) and 17.8 % (SC vs TC early hybrids). The N and PD sum of squares accounted for 0.2 and 1.3 %, respectively (Table 7). Thus, the PD sum of square contribution to the total variance in grain yield was more important relative to that of N. The variance magnitude accounted for by the remaining components ranged from 0.2 % for N x PD and G x N interaction sum of squares to 6.6 % for Rep(Environment) sum of square. The summation of the variance accounted for by the three Error terms (i.e. Error a + Error b + Error c) and replications gave approximately 38%. This proportion of the total Error variance was arbitrarily high, particularly, the magnitude of the residual, Error c (24%) ( Table 7).

Discussion
In an effort to combat the major stresses (drought, low soil N, and Striga infestation) constraining maize production in WCA, breeders have developed and commercialized extra-early and earlymaturing maize hybrids that combine high yield potentials with tolerance/resistance to the stresses and are being adopted by the farmers in the sub-region. Hybrids of both maturity groups are new to the farmers; therefore, there is a rather urgent need to recommend appropriate agronomic practices for such hybrids in the whole of WCA. In this study, the lowest density level used (66,666 plant ha -1 ) and N fertilizer rate applied (90 kg N ha -1 ) were those presently recommended for the two maturity groups. The recommendations were based on open-pollinated varieties (OPVs) in the two maturity groups developed and evaluated in the NGS zones many years ago. It was desirable to increase the rates for hybrids of these maturity groups to take advantage of the heterosis to increase production.
The results showed that the hybrids were intolerant of high plant densities and could not take advantage of higher N rates to increase production.
The results of this rather preliminary study partly confirm and refute existing knowledge, and partly open up new areas for further research on the agronomy of early and extra-early maize in the different agro-climatic zones of WCA. Our study confirmed the presence of significant differences in the performance of the hybrids both between and within maturity groups for grain yield and most other traits as earlier reported by several researchers who worked in some of the agro-ecologies used in the present study [24,25] inter alia. In another study, Oluwaranti et al. found no significant differences among varieties within maturity groups for grain yield, vegetative and flowering traits [26]. That study, which involved only OPVs, was conducted in the two seasons of the MRF agroecology used in the present study. The inconsistency of their findings with those obtained in the present study could be attributed to the difference in genetic materials and experimental design or methodology employ. Further studies are needed to resolve this conflicting situation. The significant G x E interaction effect observed for grain yield and other agronomic traits is another confirmation of existing common global knowledge of maize evaluation trials. In this study, the significant G x E interaction mean squares for grain yield was magnitudinal rather than directional; that is, the differential grain yield performance of the genotypes in all the studied environments was only in magnitude (differences in the grain yield means) and not in ranking. The early hybrids were consistently higher yielding than the extra-early hybrids in all of the agro-climatic zones, even in the Sudan savanna location, a terminal drought-prone environment. This result was rather surprising because it was expected that the extra-early hybrids would be higher yielding than the early hybrids which are later maturing and could have been a victim of the terminal drought in this location.
Partitioning the existing G x E interaction into its components is desirable when efforts are directed to releasing varieties into the ecologies of their best adaptation. Results of this study clearly indicated that extra-early hybrids are not better adapted to the Sudan savanna agro-ecology and should not be preferred to early hybrids in that ecology. In general, early hybrids are only higher yielding in all agro-ecologies, including those which have longer rainy seasons, they are not necessarily better adapted to the ecologies than the extra-early hybrids. Early maturing hybrids take a longer period to complete necessary physiological processes and grain filling before physiological maturity than extra-early maturing hybrids. The results of this study confirmed the existing knowledge of the environmental physiology of maize that extra-early varieties, including hybrids, are not necessarily more suitable than the early varieties for the Sudan savanna and, by inference, other short rainy season, terminal drought-prone environments such as the late season in the marginal forest agroecology of WCA, unless terminal drought really occurs.
Furthermore, the results of this study also, revealed that the yield performance of SC hybrids was superior to that of TC hybrids irrespective of the maturity group. The grain yield advantage of SC hybrids over the TC hybrids may be related to the variation in their genetic background and perhaps in the level of expression of heterosis. This is because SC hybrids give the maximum degree of heterosis. The higher grain yield performance of the SC hybrids may also be related to the consistent lower ear placement and reduced root and stalk lodging reported in this study. Many researchers have reported that higher ear position could increase the susceptibility to both root and stalk lodging, particularly in the extra-early, and consequently, a significant reduction in grain yield [27,28].
Therefore, the relatively lower grain yield of the TC hybrids in the present study may be linked with the higher ear placement, as well as higher root and stalk lodging consistently obtained for these hybrids in all the studied environments. Differential performance of the two maturity groups of hybrids also contributed to the significant G x E that occurred in the present study. Whereas SC hybrids were about 21% higher yielding than the TC hybrids at Zaria in the NGS, they were about obtained among four N rates (0, 150, 300 and 450 kg N ha -1 ) [32]. It was particularly striking that early and extra-early hybrids evaluated in our study did not respond to N fertilizer above 90 kg ha -1 whereas those evaluated in earlier studies, especially the more recent studies, responded to 150 kg ha -1 and higher rates. Because farmers in SSA, on average, apply less than 10 kg N ha -1 to maize, IITA  [4,33]. The present study was the first attempt at evaluating such material at N rates higher than 90 kg ha -1 . Perhaps the N response in the present study would have been different if low N rates such as 0, 30 and 60 kg ha -1 had been included in the study. Seemingly, the greater challenge to maize breeders in SSA now is to develop hybrids that would respond to high N rates for increased grain yield in commercial farms that can afford high input levels. By implication, this challenge also extends to density response, along with the non-significant N x PD interaction mean square for grain yield both of which made it impossible to determine the response surface combinations of N and PD in our study.
Lack of significant G x PD interaction effect in this study indicated that the hybrids had similar response of reduced performance in grain yield and other traits as PD increased, a confirmation of results of earlier studies on the subject-matter [7,30,32,34]. Generally, the extra-early and early maturing hybrids were intolerant of high density. Therefore, selection and development of hybrids or lines under high plant population density may be a promising strategy to improve the tolerance and adaptation of hybrid maize to higher PD. In contrast to the hybrids, PD response within locations for grain yield, yield components, and few other agronomic traits varied significantly and this was indicated by highly significant PD x E mean squares for the traits, a valid justification for extensive evaluation of density response in multiple environments in order to draw conclusion and before recommendation could be made. PD. This appears to be a general response by the plant during the growth and reproductive stages due to a reduction in photosynthate formed during these stages resulting from intense interplant competition for growth resources. Similarly, ASI value increased significantly with increased PD.
Results from other researchers have consistently shown that increased ASI is associated with increased PD due to the increased number of days to silking after anthesis [30,35,36]. The increased ASI values associated with increased PD may be related to the stress imposed on the maize plant due to intense interplant competition for light, water, and nutrients resulting from increased plant population. Fakorede and Mock reported that increased ASI is a useful indicator of density stress in maize [30] and that, by implication, could be an effective trait to use for selecting density-tolerant varieties.
The savanna locations produced a higher number of ears per plant (EPP) than the forest locations, indicating that barrenness was more pronounced in the forest than in the savanna locations. This is probably one reason; higher grain yield was obtained in the savanna than forest locations in this study. Fakorede et al. (1989), in yield trials conducted for four years, found that the yield advantage of the savanna over forest locations was due primarily to ear number. In the present study, EPP reduced and, by implication, barrenness increased with increased PD. Conversely, root lodging and stalk lodging increased with increased PD. The increased root and stalk lodgings obtained as PD increased may be attributed to stress resulting from interspecific competition for growth resources imposed by the increased plant population density. However, the magnitude of both root and stalk lodgings were larger in the forest locations than in the savanna locations implying that lodging is also largely dependent on the environment. The high repeatability estimates obtained for grain yield and most agronomic traits across the agro-climatic zones implied that the expression of the traits would be consistent under the levels of N fertilizer and plant densities.
Partitioning the total variance of multi-environment trial data into its various components among experimental factors and their interaction effects offers researchers the convenience to separate and compare the relative importance of the different variance components. In this study, E had the largest share of the total variance in grain yield but was only about 4% higher variation than the G. This is not surprising because the genotypes evaluated were improved cultivars and by implication optimization of the growing condition of the hybrids may result in a significant improvement of their grain yield. Also, the variance estimate of G was 13% higher than that of the G x E interaction. This trend of components of total variance; E ˃ G ˃ G x E has been consistently observed in earlier studies in WCA. The closeness of G to E in this study is encouraging, an indication that proper management of the E as done in this study will reduce its masking effect on the performance of the genotypes.
However, the unexpectedly large estimate of the total error variance obtained in this study suggested that more attention is still needed to minimize unexplainable sources of error in agronomic trials conducted in WCA. It is a common and routine practice of agronomists and breeders to conduct yield trials in multi-environments (locations and years) in order to identify high yielding and stable genotypes. It is, however, challenging that the effects of various management (M) practices on cultivar adaptation have not been given keen attention. Improved M practice is essential for improving grain yield particularly when the crop is managed under high plant population density.
High PD generally results in increased inter-plant competition for growth resources. Such condition can be improved with best M that involves effective control of pests, diseases, and weeds, uniformity of plant stands, and consequently, effective utilization of solar radiation and soil water, and nutrients by the maize crop. Breeding and agronomic decisions have primarily been based on G x E interaction but maize scientists seem to have neglected the significance of G x E x M.

Conclusions
In summary, the DT extra-early and early-maturing maize hybrids were genetically distinct, with the early maturing hybrids producing higher grain yield than extra-early hybrids in all the studied environments. Irrespective of the maturity group, the single-cross hybrids expressed greater yield performance relative to the top-cross hybrids primarily due to the variability in their genetic Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 27 February 2020 doi:10.20944/preprints202002.0409.v1 background, as well as the lower ear placement and reduced root and stalk lodging associated with single-cross hybrids. Grain yield advantage of the savanna locations relative to the forest locations was attributed to the number of ears per plant. The E, G, and G x E were the most important factors determining variation in grain yield. No significant difference was found for grain yield and other traits among N rates of 90 to 150 kg N ha -1 . Plant density, however, was found to affect grain yield and most of the studied agronomic traits significantly; grain yield and EPP exhibited negative linear responses, whereas ASI, PASP, EASP, RL, and SL showed positive linear responses to plant density.
The results of our investigation indicated that the genotypes were intolerant of high plant density.
We suggested that breeding programs for the improvement of early and extra-early maize germplasm for high plant density tolerance should be initiated in the sub-region. The results of this study, however, showed that 90 kg N/ha and 66,666 plants/ha were optimal for the production of extra-early and early-maturing maize hybrids across the agro-climatic zones of Nigeria.