1. Introduction
Maize (
Zea mays L.) is a major field crop whose cultivation plays a crucial role in the nutrition of the Earth’s increasing population. According to the Food and Agriculture Organization (FAO) [
1], from 2012 to 2018, the global production of maize increased by approximately 31%. Maize, just like many other field crops, usually presents significant differences in terms of performance when it is implemented in different environments. This phenomenon is recognized as the genotype × environment interaction (G × E). Therefore, this interaction is required to be assessed before introducing new cultivars that are under development for wide commercial use, as well as when introducing well-known cultivars to new environmental conditions. The main target of this evaluation was to select and separate the genotypes that present a combination of high mean performance and reliable stability across multiple environmental trials. According to Duvick [
2], the newly developed hybrids are aimed at maintaining a high and stable yield in a wide range of environments. These new hybrids can be successfully identified when tested in different environments. The adaptation of these hybrids is a long-term procedure that requires testing in many locations to be recommended for specific areas [
3,
4]. As reported by Yan et al. [
5], in multiple environmental trials, each genotype presented yield that was affected by the environment (E), the genotype (G), and their interaction (G × E).
The statistical method used to evaluate the crop yield performance, adaptability, and stability is the additive main effects and multiplicative interaction (AMMI) analysis, which can also distinguish mega-environments [
6]. AMMI analysis is now a popular statistical tool that is widely used because it can easily explain the complicated interactions of genotypes and environments [
7]. This analysis has been used in a variety of experiments with crops such as wheat [
8], maize [
9], barley [
10], and oilseed rape [
11]. In particular, Morgounov et al. [
12] used AMMI analysis to evaluate the G × E interaction for the quality of wheat grain and found that the concentration of Zn was mainly affected by the location, while Fe concentration was affected by the genotype. In another experiment, Agahi et al. [
11] used AMMI analysis to determine the effects of genotype, environment, and their interaction on the grain yield of spring oilseed rape. Similarly, Li et al. [
13] tested 20 locations and 10 genotypes in two different years to evaluate wheat yield via the use of AMMI analysis. Da Silveira et al. [
14] used AMMI analysis not only to assess the adaptability of sugarcane genotypes, but also to evaluate their phenotypic stability.
AMMI analysis combines ANOVA and PCA (principal component analysis), and the results can be used to assess yield stability in various environments and to determine the most suitable environment for the tested genotypes [
11,
15]. One crucial output of AMMI analysis is a biplot presenting the mean values of the genotypes and their relationship with the first principal component (PC) [
16,
17]. A great hybrid as a result of AMMI analysis is one that presents high mean yield along with low variability in multiple environments. In particular, that with the smallest G × E interaction is the best genotype for wide adaptation [
18].
The aim of this study was to evaluate five representative commercial maize hybrids at six locations, where maize has traditionally been cultivated for many decades, for 2 years (the combinations of year and location produced 12 environments), in order to investigate how maize yield was affected by the G × E interaction and to detect the narrow adaptations of genotypes in specific mega-environments characterized by specific soil and climatic conditions.
3. Results and Discussion
The results of ANOVA (
Table 4) indicate that, regarding grain yield (kg/ha), genotypes (GEN), environments (ENV), and their interaction (G × E) were statistically significant (
p < 0.001). According to the results, the G × E effect explained 80.36% of the variation, while GEN explained only 12.79%, with the remainder explained by ENV (6.85%). For this reason, AMMI analysis was deemed valuable to explain this high effect of G × E, in order to identify and divide the region into mega-environments, where the G × E interactions are positive. Interaction principal component 1 (IPC1), IPC2, and IPC3 were statistically significant (
p < 0.001) and the AMMI model explained 96.21% of the total G × E interaction. The result of AMMI analysis provides information on which genotype to use in which environment in order to achieve an increase in yield. Depending on the genetic material used, as well as the heterogeneity of the selected environments, similar maize grain yield experiments can present huge variation as a function of G, E, and G × E effects. Usually, the main environment effects account for more than 75% of the variation [
3,
24,
25,
26]. However, large G × E interaction effects are common in agricultural studies as it was found at the evaluation of yield of lentil [
27] and rice [
28] genotypes. When GE interaction effects are high, then the target to increase yield can be accomplished by dividing the experimental region into mega-environments and exploiting positive GE interactions of genotypes and environments. Moreover, this analysis contributes to the selection and successful introduction of new cultivars in the most suitable environments [
6].
According to the biplot of mean grain yield (kg/ha) and the first interaction principal component (
Figure 1), GEN5 had the highest mean yield (17,670 kg/ha), followed by GEN4 (17,602 kg/ha), GEN2 (16,700 kg/ha), GEN1 (15,893 kg/ha), and GEN3 (15,874 kg/ha). GEN5 was also the most stable genotype with an IPC1 score close to zero compared to the other genotypes. The IPC1 values of GEN3 and GEN4 were not very far from the zero axis, while GEN1 and GEN2 had high values of IPC1 and were considered as not stable. Concerning environments, ENV6, ENV3, and ENV12 had IPC1 values close to zero and close to the mean yield (17,102, 16,713, and 16,591 kg/ha, respectively). ENV8 had the highest mean yield (17,723 kg/ha), while ENV1 had the lowest (15,292 kg/ha).
AMMI analysis in multi-environment field trials is commonly used as a tool to evaluate maize hybrids [
9,
15,
16,
29]. The most common objective of the studies using AMMI analysis is to evaluate several genotypes in environments with similar or completely different characteristics, so as to determine the quantity and the quality of the yield. For instance, Hongyu et al. [
29] in a similar experiment evaluated the productivity of nine maize genotypes in a multi-environment experiment (20 environments) with respect to adaptability, in addition to observing the effect of the genotype × environment interaction. Similar to our study, hybrids with different FAO maturity (300–700) have been used for the evaluation of maize grain yield in Serbia, in a multi-environment experiment, using AMMI analysis (300–700) [
26]. The increasing number of studies using AMMI analysis to evaluate different genotypes in different environments, defined by soil and climatic conditions, has contributed to a better understanding of the complex genotype × environment interactions and enhanced breeding efficiency [
9,
30,
31], as well as the selection of the most suitable hybrids for the maize producers all over the world.
AMMI analysis has also been used to evaluate yield for cultivars of many other species such as wheat [
13,
32], sugarcane [
14], oilseed rape [
11], and barley [
10], as well as quality characteristics such as 1000-kernel weight in spring barley [
33] and nutritional composition of sweet potato [
34].
It is noted that the first two principal components in AMMI analysis were significant, explaining 88.46% of the variation in G × E interaction (70.25% IPC1 and 18.21% IPC2) (
Table 4). According to the biplot of the first (IPC1) and the second (IPC2) interaction principal components (
Figure 2), GEN4 had a positive interaction with ENV6, ENV8, and ENV12. When a genotype and an environment are close and have the same direction from the origin, this means that they have a positive interaction [
35]. It is important to note that the highest rain precipitation during the cultivation period was recorded in these environments. GEN2 had a positive interaction with ENV1, ENV2, ENV3, ENV5, and ENV7, where the mean monthly high temperature was over 34 °C, according to the climatic data during the cultivation period (
Table 2). Specifically, GEN2 which was located furthest from the biplot origin, contributed the largest G × E interaction for yield. Moreover, the four environments with slightly acidic pH (ENV9, ENV10, ENV11, and ENV12) had a positive interaction with all genotypes studied, except GEN2. Maize usually presents better results in neutral or slightly acidic soils; however, it can often be found in even more acidic soils. In these soils, maize can still produce increased yield, especially when cultivated with acid-tolerant cultivars [
36]. Maize plants are sensitive to high temperatures, and yield is typically decreased over 32 °C as the optimal growth ranges between 28 °C and 32 °C [
37].
Giannouli was the location with the highest mean temperature during the cultivation period (March–September) for both years (22.2 and 22.4 °C in 2017 (ENV1) and 2018 (ENV2), respectively). The lowest mean temperature was recorded at Kalamonas for both years (19.7 and 20.4 °C in 2017 (ENV9) and 2018 (ENV10), respectively). At Koutso (ENV12), as well as Nea Tyroloi (ENV6) and Kamila (ENV8), which are neighboring locations, a relatively high total precipitation (487.2 and 434.8 mm, respectively) was recorded in 2018 for the climatic conditions of Greece during the summer. The lowest precipitation during the cultivation period was recorded at Giannitsa (ENV3) in 2017 (206.6 mm).
The evaluation of environmental groups with varying climate conditions usually leads to a high G × E interaction, which means that each group has different top-yielding genotypes [
38]. Thus, when the G × E interaction component is large, no single genotype shows consistency with respect to a high grain yield [
39]. In many studies, the mega-environments derived from AMMI analysis were characterized by climatic conditions. For example, in a recent study of maize evaluation, the temperature of the environment determined the creation of the mega-environments, as well as the year [
40]. Parameters such as the quantity of rain, the soil type, and the nitrogen availability are considered of major importance in such studies, and environments sharing the same location are typically ideal for certain genotypes [
30]. In maize hybrid evaluation experiments, similar to our results, the genotypes presented specific adaptability to favorable environments [
9,
15,
41]. In a similar research in Spain, climatic conditions affected the yield of maize. Specifically, the days with mean temperature over 15 °C during the first weeks after sowing were found to have negative effect on maize grain yield, while the maximum temperature of September (up to 28.9 °C) when the plants are on the stage of seed filling was found to have a positive effect on yield [
42]. In Serbia, in a maize grain yield evaluation, high mean temperature of July affected less the yield of medium early maturing hybrids than late maturing hybrids, while the amount of precipitation (rainfed cultivation) in June was significantly correlated with grain yield [
26].
The AMMI analysis provided a table (
Table 5) with the winning genotypes for each AMMI model family in order to identify the mega-environments [
7]. Concerning AMMI model families, those to the left of the suitable AMMI model were too simple and, thus, underfitted the real signal, whereas models to the right were too complex and, thus, overfitted spurious noise (AMMI F denotes the full model or actual data). According to the first AMMI model, GEN2, GEN4, and GEN5 resulted in the most narrow adaptations, and these were the most suitable genotypes for the three mega-environments delineated (
Table 5). The first mega-environment consisted of five environments (ENV1, ENV2, ENV4, ENV5, and ENV7) in which GEN2 was the best-adapted genotype. The second mega-environment consisted of two environments (ENV3 and ENV6) with GEN5 being the best-suited genotype. The third mega-environment included five environments (ENV8, ENV9, ENV10, ENV11, and ENV12) in which GEN4 presented the best results. The genotypes and environments tested are listed in
Table 6 according to their IPC1 order. The best and worst performances in terms of IPC1 showed the opposite G × E pattern [
11]. As an example, GEN2 had a positive G × E with ENV1 and ENV2 and a negative G × E with ENV9, in which GEN1 had a positive interaction. When AMMI analysis results show narrow adaptations, it is quite common that even genotypes with high IPC scores and yield values might not be widely stable; however, they are the best-suited option in a certain environment [
15].
4. Conclusions
As an increase in yield remains a constant demand of growers worldwide, researchers have focused on providing efficient results for which genotype to use in which environment, in order to contribute to the effort to increase crop productivity. According to the first AMMI model, GEN2, GEN4, and GEN5 were the most suitable genotypes for the three mega-environments delineated. Moreover, the AMMI biplot presenting IPC2 vs. IPC1 scores for the grain yield (kg/ha) of five genotypes evaluated in 12 environments indicated that ENV9, ENV10, ENV11, and ENV12, where the soil was slightly acidic, had a positive interaction with all hybrids except GEN2. Even though slightly acidic soils are considered optimal for maize cultivation, our results show that GEN2 is not suited for soils with this characteristic. Additionally, GEN4 showed a positive interaction with ENV6, ENV8, and ENV12. These findings can be attributed to the fact that these locations recorded the highest rain precipitation (>400 mm) among the studied environments. It is important to note that the hottest environments, with the highest values in all three temperature measurements (mean monthly high, mean monthly low, and mean temperature), led to low or average yield. This study provides a valuable dataset (five representative commercial maize hybrids × six locations where maize is traditionally cultivated in Greece × two years) with results that can be exploited by breeders, seed industry companies, and farmers. AMMI analysis suggests that an increase in yield can be accomplished by dividing experimental regions into three mega-environments to exploit positive G × E interactions. The positive interaction of certain hybrids in environments with certain physicochemical soil properties and climatic conditions can be valuable for the scientific community. However, further multi-environment trials (determined by soil and climatic conditions) using more genotypes and investigating grain quality characteristics should be performed.