Genetic material

Of the 36 hard yellow corn hybrids evaluated, 29 were of double cross of yellow grain, crystalline texture and cylindrical ear, which originated from the lines of the CIMMYT; likewise, four genotypes: PM-212 (double-cross hybrid) and the experimental hybrids PM-9, PM-12 and PM-13. All the crosses of the 33 materials mentioned above were generated by the Corn Research and Social Projection Program (PIPS, for its acronym in Spanish), of the La Molina National Agrarian University (UNALM, for its acronym in Spanish), with geographical coordinates 12° 04’ 55” S and 76° 56’ 53” W, which is located at 241 masl, where the humid semi-warm climate prevails (^{SENAMHI, 2020}). The three commercial controls evaluated were: DK-5005 (trilinear, of US origin, adapted to the conditions of Peru), AG-01 (trilinear, of Brazilian origin, with a wide adaptation) and XB-8010 (double-cross hybrid, of Brazilian origin, highly productive).

Location of experiments, design and experimental unit

In 2016, experiments were established with all hybrids in the localities of La Molina (LM-2016) and Cañete (CA-2016) in the coastal area, and in Huánuco (HU-2016) and Pillco Marca (PM-2016), belonging to the low sierra region. For the 2017 agricultural cycle, the evaluation was made in the localities of La Molina (LM-2017) and Huánuco (HU-2017) and for 2018 the hybrids were evaluated only in La Molina (LM-2018). The coastal area has a humid semi-warm climate and the low Sierra region a humid temperate climate (^{SENAMHI, 2020}). Edaphoclimatic characteristics and agronomic management are shown in (Table 1).

The design used in all localities and years was a 6×6 lattice with three repetitions, the experimental unit consisted of two furrows six m long and 0.8 m wide. Three seeds per bush were sown at a distance of 40 cm and a thinning was carried out one month after sowing, leaving 64 plants per experimental plot, for a population density of 62 500 plants ha^-1.

Fertilization and trial management

In the locality of La Molina (LM-2016, LM-2017 and LM2018) and Cañete (CA-2016), the fertilization formulas 190-160-160 and 210-80-00 (N-P-K, kg ha^-1) were used, respectively, applying, at the time of sowing, all phosphorus and 50% of nitrogen and potassium (the latter element only for La Molina) 15 days after sowing (DAS) and the rest at the end of the cultivation work.

In the two localities, a preliminary irrigation and five supplemental irrigations were applied during the cultivation cycle. For Huánuco (HU-2016 and HU-2017) and Pillco Marca (PM-2016), the formula 220-115-82-21S-18Mg was used, applying 110 units of nitrogen and 100% of the other elements at 15 DAS and the nitrogen difference at 40 DAS, these localities were rainfed, with an irrigation at the beginning of sowing. In the second cultural work, for the control of weeds in the seven localities, the herbicides with active ingredient glyphosate and atrazine were applied separately, at doses of 1.5 and 1 kg ha^-1, respectively, and for the fall armyworm [Spodoptera frugiperda (J. E. Smith)] the insecticide chlorpyrifos was used with a dose of 5 kg ha^-1, which were independently diluted in 200 L of water.

Response variable

The grain yield (kg ha^-1) was calculated with the formula described by ^{Manrique (1997}):

Y=10 000A×0.971×PS×y

Where: A= area of the plot; PS= percentage of shelling (grain weight between ear weight by 100); y= yield of the experimental unit in kg adjusted to 14% of moisture. The value 0.971 corresponds to the boundary coefficient which is a constant of the experimental unit.

Statistical analysis

With the grain yield data of the 36 hybrids in the seven environments, an analysis of variance and the Gollob test were performed, using the GLM procedure of SAS^® version 9.0 (^{SAS Institute, 2012}), when significant differences between hybrids (p≤ 0.05) were detected, the Tukey mean test (p≤ 0.05) was applied, in addition, AMMI and SREG models were applied to determine the effects of GE interaction, stability and adaptation (^{Castillo et al., 2012}; Cristiano et al., 2018), with the same statistical package. Finally, a biplot graph was made for each statistical model to show the GE interaction, both biplots were generated with the first two principal components (PC1 and PC2).

AMMI analysis

Figure 1 shows that corn hybrids 26) 736×730, 15) 743×707, 14) 726×723 and 25) 729×723 were the closest to the origin of the axes; that is, they were less influenced by GE (greater stability) (^{Vargas et al., 2016}). All of them are experimental hybrids with average yields of 9.1, 9.296, 9.094 and 9.09 t ha^-1, respectively, close to the average of the corn hybrids evaluated in this research (9.188 t ha^-1).

Corn hybrids 30) Dk-5005, 31) AG-01, 36) PM-13, 27) 580×575, 32) XB-8010 and 21) 635×578 were the genotypes with the least stability (Figure 1), since they were concentrated away from the two-dimensional center. On the other hand, the hybrids 32) XB-8010, 6) 722×714 (9.213 t ha^-1) and 29) 717×714 (8.993 t ha^-1) showed low adaptability to the seven environments (Figure 1). The LM-2016 environment was the most stable and the other six environments evaluated in the present study are far from the origin (central part of Figure 1), indicating that they contributed the most to the GE interaction.

Authors such as (^{Vargas et al., 2016}) indicated that in environments with angles < 90º, genotypes will preserve a similar spatial distribution, which occurred between environments LM-2016, LM-2017 and LM-2018. A similar situation with angles < 90º occurred between the environments HU-2016, PM-2016 and CA-2016. On the other hand, environments with angles greater than 90º do not order genotypes in the same way, as happened with HU-2016 and HU-2017, LM-2016 or LM-2018 and CA-2016 (Figure 1).

Figure 1 AMMI biplot with the first two principal components (PC1 and PC2) of the average yield of 36 hard yellow corn hybrids in seven environments in Peru. 

The environments that are located at an angle of 180º order the corn genotypes in the opposite way, which makes it difficult to select through these environments because they are of contrasting conditions (^{Yang et al., 2009}; ^{Kandus et al., 2010}), as happened in the environments LM-2016, LM-2017 and LM-2018 against HU-2016 (Figure 1), possibly because the first corn materials are from the coast and the other from the Sierra of Peru.

The vectors with the groups of the environments LM-2016, LM-2017 and LM-2018, as well as the environments HU-2016, PM-2016 and CA-2016 are too close to each other, which may be due to the edaphoclimatic similarity (fertilization, soil type, precipitation and temperature with minor differences). Thus, in the first case (the environments of La Molina), it is the same site, but evaluated in a different year (Table 1); that is, the fact that the environments are very close to each other in Figure 1 and in a similar quadrant (similar direction of the vectors) means that they are very similar to the GE. In the second case, the three environments belonged to the same year of evaluation (2016), which could be a determining factor for the proximity between them; results similar to the findings of the present work were found by (^{Castillo et al., 2012}; ^{López-Morales et al., 2019}) under the same conditions.

SREG Analysis

The corn hybrid 30) Dk-5005 was the one with the highest yield (p≤ 0.05); while the hybrid 5) 515×714 had the lowest yield. The Dk-5005 hybrid was not only the one with the highest yield but also the one with the greatest adaptability and stability in the seven environments (Figure 2).

Other stable materials were the hybrids: 21) 635×578, 2) 685×684, 7) 532×531, 18) 513×531, 28) 742×737, 26) 736×730 and 12) 694×691, all experimental corns. Also, in environments, the length of the vector indicates the variability in yield explained in each environment and vice versa (^{Crossa et al., 2015}).

On the other hand, six mega-environments (region with homogeneous environment for a crop species) were found (Figure 2) and where each mega-environment (between red lines in Figure 2) will locate hybrids with a high yield at the vertices that will form a polygon and all environments that are outside this polygon allow discrimination between hybrids (^{Yan et al., 2011}; ²⁰¹⁶).

Figure 2 SREG biplot with the first two principal components (PC1 and PC2) of the average yield of 36 hard yellow corn hybrids in seven environments in Peru. 

This happened with the mega-environment where the corn hybrid 30) Dk-5005 showed the highest yield between genotypes (11.524 t ha^-1) and included the environments LM-2016 (the most stable of the four, but the only one that did not discriminate between genotypes), LM-2017, LM-2018 (the three environments with the best yields according to the Tukey test: 9.134, 12.65 and 10.337 t ha^-1, respectively) and HU-2017 (8.062 t ha^-1).

Other materials with good response in the same mega-environment, with the highest yields among the materials evaluated, were the corn genotypes 31) AG-01 and 27) 580×575 (experimental corn). Both Dk-5005 and AG-01 are trilinear hybrids with a high frequency of genes due to their genetic constitution of three lines, producing a high adaptive capacity for several areas, as indicated by the results of ^{Chura and Huanuqueño (2014}); ^{López-Morales et al. (2019}) with the materials of three lines in different regions of Peru.

The sites PM-2016 and CA-2016 were located in a mega-environment, but in this one there was no corn hybrid that excelled in yield, even though the PM-2016 environment discriminated between hybrids. Such behavior could be due to the large edaphoclimatic differences between the two environments, especially in the amount of rainfall, Table 1. Between the first and second quadrant of the biplot of Figure 2 is another mega-environment where the only environment was HU-2016, for which the corn hybrid 32) XB-8010 was the ideal.

The other three mega-environments did not include any environment, probably because the seven environments manifested yield averages similar to a group of hybrids. The ideal hybrids in each of these mega-environments were: 5) 515×714, double-cross hybrid with the lowest yield (8.045 t ha^-1) and genotypes 19) 739×737 and 10) 725×723, located in the last places of yield (with 8.069 and 8.555 t ha^-1, respectively), together in a single mega-environment (Figure 2).

In general, in Figure 2 the hard yellow corn hybrids, located at the right vertices, are those with the highest grain yield (Dk-5005 and AG-01) and are intimately linked to adjacent environments. The localities were grouped into three of the six mega-environments. Hybrids located between quadrants II and III away from the lines of the localities showed poor yield performance, as they were far away from the localities.

The corn hybrids studied in the present work consistently had high positive scores in PC1, which means high production, and decreases in absolute scores of PC2 are indications of high stability. Similarly, environments with low absolute values in PC2 have more representativeness, and high positive evaluations in PC1 indicate greater discriminatory capacity of hybrids in terms of the main genotypic effect (^{Yan et al., 2011}; ²⁰¹⁶).