Introduction
Durum wheat (Triticum turgidum L. subsp. durum) is one of the first domesticated food crops; is being the most cultivated cereal crop in the Mediterranean environments, traditionally grown under rainfed conditions, with about 17 million ha cultivation worldwide and 38.1 million tonnes production in 2019 (FAO, 2019). The Mediterranean environments are usually characterized by large and unpredictable fluctuations in rainfall and temperature within and among cropping seasons, resulting in high genotype × environment (GE) interaction and crop production instability, which complicate the interpretation of performance genotypes in multi-location trials from year to year (Mohammadi et al., Reference Mohammadi, Sadeghzadeh, Poursiahbidi and Ahmadi2021).
Drought as the most important abiotic stresses constraints crop production and productivity worldwide and significantly affecting world’s food security particularly in the arid and semi-arid regions (Mohammadi, Reference Mohammadi2018). Various reports from around the world indicate that drought stress plays a major role in yield reduction of wheat. According to FAO (2018), global wheat production in 2018 was predicted to decline by 2.7% which is based on predictions of climate change. A meta-analysis of 60 published reports indicated that drought reduced wheat yields by an average of 27.5% (Zhang et al., Reference Zhang, Zhang, Cheng, Jiang, Zhang, Peng, Lu, Zhang and Jin2018), and a similar study, which included peer-reviewed articles from 1980 to 2015, showed decreases of 20.6% (Daryanto et al., Reference Daryanto, Wang and Jacinthe2016). The impact of drought stress on total yield varies with the region, crop and cultivar and the occurrence of other stresses such as high temperatures (Mohammadi et al., Reference Mohammadi, Sadeghzadeh, Armion and Amri2011a). Severe droughts in Australia resulted in yield reductions of five major field crops including wheat by 25–45% compared to the years with optimum rainfall (Madadgar et al., Reference Madadgar, AghaKouchak, Farahmand and Davis2017). Iran is periodically affected by severe droughts every 5–7 years. Our previous studies also showed up to 59% reduction, depending on drought stress intensity, in rainfed durum wheat production (Mohammadi, Reference Mohammadi2016; Mohammadi et al., Reference Mohammadi, Armion, Kahrizi and Amri2010). In the last two cropping seasons (2020–2021 and 2021–2022), due to severe droughts, the long-term average rainfall of the country from 250 mm annually reached to about 150 mm, which caused about 50 to 70% reduction in rainfed wheat production.
Understanding different morphological, agronomic and physiological responses to drought stress may improve drought tolerance in wheat (Shirvani et al., Reference Shirvani, Mohammadi, Daneshvar and Ismaili2023). In addition, different crop managements and development of new cultivars are recommended to address the main challenges of crop adaptation to abiotic stresses such as drought (Carvalho et al., Reference Carvalho, Azam-Ali and Foulkes2014; Waqas et al., Reference Waqas, Wang, Zafar, Noor, Hussain, Azher Nawaz and Farooq2021; Yang et al., Reference Yang, Liu, Li, Wang, Yin and Deng2021). Therefore, it is important to apply methods that measure the adaptability and stability to identify the best genotypes under variable environmental conditions. Various breeding methods have been applied to improve drought tolerance in crop species through the integration of morphological and physiological drought tolerance systems of resistant genotypes (Chowdhury et al., Reference Chowdhury, Hasan, Bahadur, Islam, Hakim, Iqbal, Javed, Raza, Shabbir, Sorour, El Sanafawy, Anwar, Almari, EL Sabagh and Isalm2021; Khadka et al., Reference Khadka, Earl, Raizada and Navabi2020; Zhang et al., Reference Zhang, Zhang, Cheng, Jiang, Zhang, Peng, Lu, Zhang and Jin2018). For example, for the field evaluation of germplasms, various morphological, phenological and physiological traits have been developed by international institutions (ICARDA and CIMMYT) and their national partnerships institutions around the world, which play an important role in the adaptation of a particular crop to changing environmental conditions. Among these traits, days to heading and maturity, plant height, 1000-kernel weight and grain yield are among mostly recommended traits for field evaluation of germplasms particularly under rainfed conditions (Mohammadi et al., Reference Mohammadi, Armion, Sadeghzadeh, Amri and Nachit2011b; Bassi and Sanchez-Garcia, Reference Bassi and Sanchez-Garcia2017; Gerard et al., Reference Gerard, Crespo-Herrera, Crossa, Mondal, Velu, Juliana, Huerta-Espino, Vargas, Rhandawa, Bhavani, Braun and Singh2020).
Due to frequent changes in amount and distribution of rainfall during the crop growth from year to year in the Mediterranean environments, it is crucial to assess the response of promising genotypes in breeding programmes in multi-year trials and sites differing in water regime conditions, before recommendation of any genotype for release. This condition provides some useful information on tolerance/resistance of promising genotypes to drought for a specific environment (Fernandez, Reference Fernandez1992; Mohammadi, Reference Mohammadi2016; Negisho et al., Reference Negisho, Shibru, Matros, Pillen, Ordon and Wehner2022). For this purpose, several drought tolerance indices, which are based on performance in stress and non-stress conditions, have been used to increase the efficiency of selection and screening of genotypes grown under drought stress conditions. Fischer and Maurer (Reference Fischer and Maurer1978) proposed the stress susceptibility index (SSI) to measure the susceptibility/resistance of genotypes to stress conditions. Rosielle and Hamblin (Reference Rosielle and Hamblin1981) proposed the index of tolerance (TOL) as the difference in performance of a genotype in the both conditions and the index of mean productivity (MP) as the mean performance of a genotype in both stress and non-stress conditions. Bouslama and Schapaugh (Reference Bouslama and Schapaugh1984) proposed yield stability index (YSI) to evaluate stability of genotypes in both stress and non-stress conditions. Bidinger et al. (Reference Bidinger, Mahalakshmei and Rao1987) presented a drought response index (DRI) to select genotypes that are tolerant or susceptible to drought. This index corrects yield under drought conditions for variation in flowering date and potential yield under irrigated conditions, thus ensuring that selected genotypes will have drought tolerance characteristics. Fernandez (Reference Fernandez1992) introduced the stress tolerance index (STI) and geometric mean productivity (GMP) to identify genotypes based on their performance in contrasting environments. Gavuzzi et al. (Reference Gavuzzi, Rizza, Palumbo, Campaline, Ricciardi and Borghi1997) suggested the yield index (YI) as a measure of the performance of a genotype under stress condition relative to overall mean under stress condition. These indices have been widely used for screening drought-tolerant genotypes in wheat (Ayed et al. Reference Ayed, Othmani, Bouhaouel and Teixeira da Silva2021; Mohammadi, Reference Mohammadi2016; Rana et al., Reference Rana, Singh, Dhiman and Chaudhary2014; Song et al., Reference Song, Liu, Goudia, Chen and Hu2017).
Multi-environment trial (MET) analysis, in most cases, is conducted based on only a single trait, such as grain yield. However, the reliability in recommending genotypes is increased when several traits are considered. For this purpose, recently a technique for MET analysis (METAN) introduced by Olivoto et al. (Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a), which allows the evaluation of genotypes based on multiple traits into a single index, namely multi-trait stability index (MTSI), could provide a unique selection process.
MTSI is a useful technique for selection of genotypes based on multiple traits because it provides a robust and simple selection process (Olivoto et al., Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a, Reference Olivoto, Lúcio, Silva, Sari and Diel2019b). This method has been successfully applied to select high-yielding genotypes in variable environments in several crop species (Al-Ashkar et al. Reference Al-Ashkar, Sallam, Almutairi, Shady, Ibrahim and Alghamdi2023; Norman et al., Reference Norman, Agre, Asiedu and Asfaw2022; Olivoto et al., Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a; Zuffo et al., Reference Zuffo, Steiner, Aguilera, Teodoro, Teodoro and Busch2020). This study was undertaken to evaluate 25 durum wheat genotypes under rainfed and irrigated conditions during three cropping seasons to (i) identify the stability traits associated with the mean performance and (ii) identify genotypes with the best performance and drought tolerance in high- and low-rainfall environments.
Materials and Methods
Experimental site
The experiments were conducted in Sararood dryland agricultural research station (longitude 37o19'N; latitude 47o17'E; altitude 1351 m), Kermanshah, Iran, representative of moderate cold climate in rainfed wheat breeding programme. The site is characterized by annually average minimum, maximum and mean temperature of 3.5, 19.1 and 11.5 °C, respectively, and 60–100 days with temperatures below zero annually and average humidity of 51%. The average long-term annual precipitation is estimated to 435 mm, consisting of 95% rain and 5% snow. The soil at the site was clay loam.
Experimental material, layout, design and trial management
The study was conducted by using 25 durum wheat genotypes (Table 1) including 23 elite breeding lines and two national checks under both rainfed and supplemental irrigation conditions for three cropping seasons (2019–2020, 2020–2021 and 2021–2022), resulting in six environments. The experiments serve as part of 27th elite regional durum wheat yield trials (27th ERDYT) and represent the final stage of the national rainfed durum wheat breeding program. In each environment, the experimental design was a randomized complete block design (RCBD) with three replications. The linear mixed model used to calculate the variance components of the studied traits adopted the following equation:
$${Y_{ijk}} = \mu + {\alpha _i} + {\tau _j} + {\left( {\alpha \tau } \right)_{ij}} + {\gamma _{jk}} + {\varepsilon _{ijk}}$$
where Y ijk is the response variable (e.g., grain yield) of the kth block of the ith genotype in the jth environment (i = 1, 2,…, g; j = 1, 2,…, e; k = 1, 2,…, b); μ is the grand mean; α i is the main effect of the ith genotype; τ j is the main effect of the jth environment; (αt) ij is the interaction effect of the ith genotype with the jth environment; γ jk is the effect of the kth block within the jth environment and ε ijk is the random error which was assumed to be normally and independently distributed with a mean of 0 and a variance of σ 2.
Durum wheat genotypes evaluated across rainfed and irrigated conditions during three cropping seasons (2019–2020, 2020–2021 and 2021–2022)
The size of each experimental plot was 7.2 m2 (6 rows, 6 meter-long, 20 cm row spacing). The sowing density was 400 grains per m2. The fertilizers used were 50 kg ha−1 N and 50 kg ha−1 P2O5 as basal application at sowing. Weeds were controlled and managed by herbicides and hand weeding in each environment. The irrigated trials were conducted to identify high-yielding potential genotypes in durum breeding program. In irrigation experiments, two irrigations, each with 30 mm, were applied using sprinkler system during heading and grain-filling stages to mitigate terminal drought stress effects.
Data collection
In each experiment, days to 50% heading (DHE) and physiological maturity (DMA), plant height (PLH) measured from the soil level to tip of spike (excluding the awns), 1000-kernel weight (TKW) and grain yield (YLD) were recorded. Grain yield was measured as kg per plot, and then converted to yield per hectare (kg/ha).
Eight drought tolerance and susceptibility indices were calculated for each genotype based on the grain yield under drought and irrigated environments, according to the following formula, to differentiate the drought tolerant and susceptible genotypes.
-
1)
${\rm{Stress}}\;{\rm{tolerance}}\;{\rm{index}}\;\left( {{\rm{STI}}} \right) = {{\left( { Ys} \right)\left( {Yp} \right)} \over {{{\left( {\overline Yp} \right)}^2}}}$
(Fernandez, Reference Fernandez1992) -
2)
${\rm{Geometric\;mean\;productivity\;}}\left( {{\rm{GMP}}} \right) = \;\sqrt {\left( {Ys} \right)\left( {Yp} \right)} $
(Fernandez, Reference Fernandez1992) -
3)
${\rm{Mean}}\;{\rm{productivity}}\;\left( {{\rm{MP}}} \right) = \;{{\left( {Ys + Yp} \right)} \over 2}$
(Rosielle and Hamblin, Reference Rosielle and Hamblin1981) -
4)
${\rm{Tolerance\;index\;}}\left( {{\rm{TOL}}} \right)$
= Yp-Ys (Hossain et al., Reference Hossain, Sears, Cox and Paulsen1990) -
5) Stress susceptibility index (SSI) =
${{\left[ {1 - \left( {{{Ys} \over {Yp}}} \right)} \right]} \over {1 - {\rm{SI}}}};{\rm{stress}}\;{\rm{intensity}}\;\left( {{\rm{SI}}} \right)\;$
=
$\left[1\!-\!\left( {\overline Ys} \right)/\left( {\overline Yp} \right)\right]$
(Fischer and Maurer, Reference Fischer and Maurer1978) -
6) Yield stability index (YSI) =
${{Ys} \over {Yp}}$
(Bouslama and Schapaugh, Reference Bouslama and Schapaugh1984) -
7) Yield index (YI) =
${{Ys} \over {\overline Ys}}$
(Gavuzzi et al., Reference Gavuzzi, Rizza, Palumbo, Campaline, Ricciardi and Borghi1997) -
8) Drought response index (DRI)
$ = {{\left( {{Y_S} - {Y_{est}}} \right)} \over {S{E_{Yest}}}}$
;
${Y_{est}} = a + b{Y_p} + cD{F_P}$
(Bidinger et al., Reference Bidinger, Mahalakshmei and Rao1987)
where Y
s
and Y
p
stand for the mean yields of each genotype under drought and irrigated conditions, respectively;
$\overline Ys$
and
$\overline Yp$
, respectively, represent mean yields of genotypes under drought and irrigated conditions.
${Y_{est}}$
stand for estimated grain yield of specific genotype under rainfed condition;
$S{E_{est}}$
is standard error of estimated grain yield of all genotypes under rainfed condition;
$D{F_P}$
is days to heading under irrigated condition; and a, b and c are regression parameters.
Statistical analysis
The data collected on measured traits from each environment were subjected to combined ANOVA using a mixed linear model. The genotype effect was treated as fixed, and the environment, GE interaction and replications were treated as random factors.
The MTSI was applied to rank and select genotypes in the two levels of agronomic traits and drought tolerance indices. In our study, the genotype selection aimed at selecting genotypes with higher values (positive gains) for grain yield, TKW and plant height and lower values of phenological traits (DHE and DMA). In the case of drought tolerance indices, the genotype selection aimed at higher values of STI, GMP, MP, YI, YSI, DRI and lower values of SSI and TOL. The analyses were done according to Olivoto and Lúcio (Reference Olivoto and L’ucio2020) using the “metan” package in R software. 15% of the top genotypes with the lowest MTSI score were selected and marked with red colour in the MTSI plot. The MTSI index using the investigated traits/indices for 25 genotypes was calculated according to the following formula (Olivoto et al., Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a):
$$MTS{I_i} = \mathop \sum \limits_{j = 1}^f {\left[ {{{\left( {{F_{ij}} - {F_j}} \right)}^2}} \right]^{0.5}}$$
where the MTSI is the multi-trait stability index for the ith genotype, F ij is the jth score of the ith genotype and F j is the jth score of the ideotype genotype. The genotype with the lowest value of MTSI is closer to the ideotype, showing the high mean performance and stability based on the investigated variables.
A heat map correlation analysis was calculated among drought tolerance indices and genotypic mean yields under rainfed and irrigated conditions to study the relationships between the indices and mean yields in the both condition. Principal component analysis (PCA) was applied to understand the interrelations among drought tolerance indices and the interaction between genotypes and indices in each and across years.
Results
Weather conditions
The three cropping seasons differed in amount and monthly rainfall distribution, which caused contrasting growing conditions and therefore a range in yield potential under rainfed conditions (Fig. 1). However, rainfall varied between cropping seasons, and the genotypes were exposed to drought stress. The 2020–2021 and 2021–2022 cropping seasons were characterized by abnormally low rainfall of 317.5 and 227.5 mm, respectively, relative to optimal cropping season 2019–2020 with 518.8 mm rainfall, which was relatively higher than the long-term average (435 mm). In contrast to rainfall, no marked variation in temperatures was observed across cropping seasons, although the drier seasons were also warmer particularly in heading and grain filling period which coincided with high terminal drought stress (Fig. 1). The growing conditions in 2020–2021 and 2021–2022 due to severe drought had negative effects on performance of experiments, resulted in abnormal seasons, but 2019–2020 cropping season was found to be normal due to higher rainfall than the long term, although insufficient rainfall was recorded during May and June, the accumulated soil moisture content and optimum temperature during grain filling period reduced terminal drought stress in this cropping season.
Monthly rainfall distribution and average temperature in Sararood dryland agricultural research station during three cropping seasons.
Variance components and overall performance
The results of combined analysis of variance for traits studied showed that the environment, genotype and GE interaction effects were significant for all traits studied. The environment was the main source of variation for all traits except TKW (83% for YLD; 76% for PLH; 65% for DHE and 55% for DMA). For grain yield and plant height, the GE interaction variance was higher than genotypic variance, showing high GE interaction for these traits (Fig. 2). The genotypic variance for TKW and phenological traits was higher than GE interaction variance, showing high genotypic variation.
Proportion of the phenotypic variance for investigated traits of 25 durum wheat genotypes across six environments. DHE: days to heading; DMA: days to maturity; PLH: plant height; TKW: 1000-kernel weight; YLD: grain yield.
The overall mean yield was 3078 ± 106.7 kg/ha, with ranging from 1173 to 5913 kg/ha across all environments. The highest overall yield (4592 kg/ha) was recorded in 2019–2020, while in 2020–2021 and 2021–2022 it was 2052 and 1569 kg/ha, respectively, showing a high yield reduction of 55% and 66% compared to 2019–2020. These reductions were mostly due to severe droughts in 2020–2021 and 2021–2022. The grand mean for TKW was 37 ± 0.3 g and varied from 27.8 to 49.8, whereas for plant height, overall mean was 72 ± 1.2 cm ranged from 45 to 104 cm; for days to heading the mean value was 118 ± 0.3 days varied from 109 to 127 days; and for days to maturity the mean value across environments was 160 ± 0.3 ranged from 152 to 168 days (data not shown).
Factor analysis for agronomic traits
Table 2 presents the eigenvalues, explained variance, factorial loadings after varimax rotation, and communalities obtained in the factor analysis for 25 durum wheat genotypes in each and across environments. In rainfed condition during 2019–2020, the first two factors with eigen values >1 were retained and accounted for 69.4% of the total variance (Table 2). After varimax rotation, the average communality (h) was 0.69, with values ranging from 0.49 (grain yield) to 0.82 (plant height). The first factor (FA1) had highest loadings for days to heading (DHE), 1000-kernel weight (TKW) and grain yield (YLD), and second factor (FA2) represented the highest contribution for the traits days to maturity (DMA) and plant height (PLH).
Eigenvalues (EV), explained variance (Var%), factorial loadings after varimax rotation and communalities (h) obtained in the factor analysis for 25 durum wheat genotypes in each rainfed irrigated environments during single and across three cropping seasons
FA1, FA2 and FA3 represented for first, second and third factors. DHE: days to heading; DMA: days to maturity; PLH: plant height; TKW: 1000-kernel weight; YLD: grain yield. The bold values represent the traits with high contribution to each factor.
Under irrigated condition in 2019–2020, the first two factors were retained and captured for 76% of variation (Table 2). The average communality (h) was 0.68, with values ranging from 0.31 for TKW to 0.87 for DHE. The FA1 represented the highest contribution for traits DHE, DMA and TKW, while FA2 represented the highest contribution for traits PLH and YLD, as they had highest loadings on their correspondence factor.
In the case of rainfed condition in 2020–2021, the first two factors were retained, accounting for 62.8% of variance. The average communality was 0.63, with values ranging from 0.45 for DMA to 0.81 for PLH. The FA1 represented the highest contribution for the traits PLH and YLD, while the FA2 represented the highest contribution for the traits TKW, DMA and DHE.
Under irrigated condition, the first three factors with eigen values greater >1 were retained and explained 85.6% of the total variation (Table 2). The average communality (h) was 0.85; with values ranging from 0.80 for DHE to 0.93 for PLH. The FA1 represented the highest contribution for the traits DMA and DHE, while FA2 associated with YLD and TKW, and third factor (FA3) correlated with plant height. The average communality accounted for 77% of all the genetic variability in the dataset with highest (h = 0.93) for plant height and lowest (h = 0.80) for days to heading.
In 2021–2022 for the rainfed environment, the first two factors captured 68.3% of variance, with average communality of 0.682 varied between 0.43 for TKW and 0.82 for PLH. The FA1 most associated with phenological traits (DHE and DMA) and TKW, whereas FA2 most correlated with PLH and grain yield. Under irrigated condition, the first two factors explained 65% of total variation (Table 2). The average communality was 0.65 with values ranging from 0.48 for DM to 0.74 for PLH. The first factor represented the highest contribution for PLH and TKW, while the second factor correlated with YLD, DHE and DMA.
When taking all rainfed environments into consideration, the first two factors captured 67.7% of variance, with average communality of 0.68; varied from 0.45 for DMA and 0.87 for PLH (Table 2). The first factor associated most with phenological traits and TKW, while the second factor correlated with PLH and grain yield. Across irrigated environments, the first three factors with eigen values greater than unit were retained and cumulatively accounted for 90.2% of the total variation (Table 2). The average communality was 0.90 with values ranging from 0.87 for DMA to 0.90 for DHE. The first factor associated with phenological traits, while second factor represented the highest contribution for PLH and TKW, and the third factor most associated with grain yield.
Multi-Trait stability index and genotype selection
Figure 3 shows the genotype ranking by MTSI plot method described by Olivoto et al. (Reference Olivoto, Lúcio, Silva, Sari and Diel2019b), which allows selection of the best genotypes by employing information from a five drought-adaptive traits in each and across environments. When considering the index for rainfed condition in 2019–2020 (Fig. 3a), the breeding lines G20, G25, G24 and G16 with lowest MTSI values of 1.017, 1.548, 1.581 and 1.653 were identified as most stable genotypes among the 25 durum genotypes. The MTSI value of 1.653 indicates cut-point (red circle in Fig. 3a), considering the selection intensity. The breeding lines G13 and G22 also were near to the cut-point and could present interesting features. Thus, in future studies, it would be interesting to investigate the performance of these genotypes that are very close to the cut-point.
Genotype ranking in ascending order for the MTSI based on five drought-adaptive traits in each and across environments (2019–2022). The selected genotypes based on MTSI are shown in red. The central red circle represents the cut point according to the selection pressure of 15%. RF and IR refer to rainfed and irrigated conditions, respectively.
Figure 4 shows the strengths and weaknesses view of the stable genotypes identified by MTSI plot in each and across environments. The factors that contributed the most were placed towards the centre, while those that contributed less were drawn near the plot edge. Under rainfed condition in 2019–2020, the first factor had the highest contribution for G16, and thus is characterized for high values of TKW, DHE and YLD, showing G16 best performing based on these traits; while FA2 had the highest contribution for G25, followed by G20 and thus expressed for highest values of PLH and DMA. The selection differential (the difference between the population mean and the mean of the selected genotypes) was positive for agronomic traits (PLH, TKW and YLD) and negative for phenological traits (DHE and DMA), suggesting that the method is efficient in selecting high performing and stable genotypes with early heading and maturity traits. The mean selection differential for the investigated traits in trial was 2.31%, being the lowest for phenological traits (DHE and DMA) and the highest for grain yield (6.59%), followed by TKW (4.84%).
The strengths and weaknesses view of the stable genotypes identified in single and across three consecutive growing seasons, shown as the proportion of each factor on the estimated MTSI. Smaller proportions explained by a factor (closer to the external edge) indicate that the trait within that factor is closer to the ideotype. The dashed line shows the theoretical value if all the factors contributed equally. RF and IR refer to rainfed and irrigated conditions, respectively.
Figure 3b presents the genotype ranking using the MTSI for the traits studied under irrigated condition in 2019–2020. The breeding lines G20, G18, G17 and G24 with lowest MTSI values of 1.47, 1.63, 1.789 and 1.913, respectively, were selected. The red circle showing cut-point with MTSI = 1.913, considering the selection intensity. Breeding line G6 also with MTSI = 2.187 was near to red circle and could present interesting features. Among the four selected genotypes, breeding line G24 with highest contribution to MTSI was related to FA1 (Fig. 4b). The DHE, DMA and TKW showed highest contribution to FA1 (Table 2), showing G24 expressed for highest values of these traits. Breeding line G20, contributed most to MTSI in regard to FA2, and PLH and YLD showed highest contribution to FA2, indicating breeding line G16 expressed best performance based on yield and plant stature. Thus, under irrigated condition, the objective can be aimed at improving the genotypes with highest mean yield and high plant stature, earliness and average TKW (average lodgings on FA1). Under irrigated condition, the selection differential was positive for agronomic traits (PLH, TKW and YLD) and negative for phenological traits (DHE and DMA), suggesting that the method is efficient in selecting high performing and stable genotypes with early heading and maturity traits. The mean selection differential for the investigated traits was 5.14%, being the lowest for phenological traits (DHE and DMA) and the highest (16.4%) for grain yield, followed by PLH (6.56%) (Table 3).
Estimates of the original mean (Xo), mean of the selected genotypes (Xs), selection differential (SD) and percentage selection differential (SD%) based on multi-trait stability index applied on five investigated traits for 25 durum wheat genotypes under rainfed and irrigated environments during three cropping seasons
DHE: days to heading; DMA: days to maturity; PLH: plant height; TKW: 1000-kernel weight; YLD: grain yield.
Figure 3c shows the genotype ranking based on MTSI under rainfed condition in 2020–2021. The genotypes G20, G3, G16 and G2 expressed lowest values for MTSI (1.417, 1.419, 1.470 and 1.901, respectively) and can be considered as most stable genotypes. The genotype G23 with MTSI equal to 1.981 also was near to the cut-point (1.901) and could present interesting features. Among the stable genotypes, G16 and G20 expressed for highest contribution to MTSI in regard to FA2 (Fig. 4c), being the highest correlation with TKW and YLD.
The selection differential was positive for TKW, PLH and YLD and negative for days to heading and maturity. The mean selection differential was 4.734%, with lowest for DHE (−1.13) and the highest (12.1%) for plant height, followed by grain yield (4.80%) (Table 3). Under irrigated condition in 2020–2021 (Fig. 3d), breeding lines G25, G20, G6 and G21 were identified as stable genotypes for traits studied. The MTSI values for these genotypes was 1.03, 1.945, 2.011 and 2.128, respectively. Breeding lines G24 and G5 with MTSI values of 2.142 and 2.199 were near to the cut-point (2.128), and they can provide interesting features. Among the four stable genotypes, breeding lines G6, G21 and G20 showed highest contribution to FA2 (Fig. 4d), being the highest correlation with TKW and grain yield, while breeding line G25 contributed most to FA1 being the highest association with phenological traits (DHE and DMA). Breeding line G25 also showed high contribution to FA3 represented for the highest value of plant height (Table 2). The selection differential was positive for PLH, TKW and YLD and was negative for phenological traits (DHE and DMA), suggesting that the method is efficient in selecting high performing and stable genotypes with early heading and maturity traits. The mean selection differential was 2.678%, with lowest (−1.38) for DHE and the highest (5.89%) for TKW followed by grain yield (4.80%) (Table 3).
Figure 3e shows ranking of genotypes for MTSI under rainfed condition in 2021–2022. The genotypes G6, G7, G8 and G25 were considered as most stable genotypes as they expressed lowest values for MTSI (1.407, 1.478, 1.532 and 1.650, respectively). The genotype G23 with MTSI equal to 1.824 was also near to the cut-point (1.650) and can be regarded for further evaluations. Among the stable genotypes, G25 expressed highest contribution to MTSI which related to FA1 (Fig. 4e), showing this genotype expressed for highest phenological traits and TKW, while G6 exhibited highest contribution to MTSI and was related to FA2, indicating that G6 can be characterized for highest TKW and YLD under rainfed condition. The selection differential was positive for TKW, PLH and YLD and negative for days to heading and days to maturity. The mean selection differential was 5.36%, with lowest for DMA (−0.554) and the highest (9.91%) for plant height, followed by TKW (9.15%) (Table 3).
Under irrigated condition in 2021–2022 (Fig. 3f), genotypes G6, G25, G2 and G24 with MTSI values ≤ cut-point (MTSI = 1.92) were selected as stable genotypes. The genotypes G19 and G20 with MTSI values (1.999, 2.096) near to the cut-point, could also be regarded in breeding programme. Genotypes G6 and G25 had strengths related to FA2 and displayed high traits values for phenological traits and mean yield (Fig. 4f and Table 2). Under irrigated condition, the selection differential was positive for PLH, TKW and YLD and was negative for phenological traits (DHE and DMA), suggesting that the method is efficient in selecting high performing and stable genotypes with early heading and maturity traits. The mean selection differential was 3.647%, with the lowest (−0.715) for DMA and the highest (9.54%) for PLH followed by TKW (8.41%) (Table 3).
Across three rainfed environments during 2019–2022 (Fig. 3g), among 25 durum genotypes evaluated, breeding lines G6, G19, G20 and G2 were identified as high performing genotypes for investigated traits. According to the strengths and weaknesses view of the genotypes, G2 showed most contribution in regard the first factor which associated with phenological traits and TKW (Fig. 4g), while breeding lines G6, G20 and G19 exhibited the highest contribution to second factor and thus characterized for mean yield and plant stature (Fig. 4g; Table 2). Across rainfed environments, the selection differential was positive for TKW, PLH and YLD; and negative for phenological traits. The mean selection differential was 2.79%, with lowest for DMA (−0.829) and the highest (8.62%) for plant height (4.80%) (Table 3).
Across three irrigated environments during 2019–2022 (Fig. 3h), breeding lines G20, G6, G24 and G18 were selected as high performing and stable genotypes for investigated traits. According to the strengths and weaknesses view of genotypes, the G24 and G18 have strength related to first factor and characterized for phenological traits (Fig. 4h), while G18 also have strength related to second factor represented for high plant height and TKW, and genotypes G20 and G6 were associated with the third factor and are characterized by high yield. Across irrigated environments, the selection differential was positive for TKW, PLH and YLD and negative for days to heading and maturity. The mean selection differential was 3.429%, with lowest value for days to heading and highest for grain yield (Table 3).
Considering genotypic performance across environments (Fig. 3i), breeding lines G20, G6, G25 and G18 selected as high performing and stable genotypes. Based on the strengths and weaknesses view of genotypes, the G25 and G18 have strength related to first factor and characterized for phenological traits (Fig. 4i), while G18 showed most strength related to second factor represented for high TKW and plant stature, and genotypes G6 and G20 were associated with the third factor and are characterized by high yield.
Selection gains for agronomic performance
The multitrait selection resulted in a success rate in selecting traits with desired selection differentials (SD) in investigated environments (Fig. 5), as these traits are most applied in international institutes, e.g., ICARDA and CIMMYT and national partnerships institutes around the world for germplasm evaluations. The four selected genotypes with 15% pressure selection (ranked by MTSI plot) under rainfed condition in 2019–2020 were G20, G25, G24 and G16 (Fig. 3a) and in irrigated condition were G20, G18, G17 and G24 (Fig. 3b). In 2020–2021, the selected genotypes under rainfed condition were G20, G3, G16,and G2 (Fig. 3c) and under irrigated condition were G25, G6, G6 and G21 (Fig. 3d). In 2021–2022, the selected genotypes under rainfed condition were G6, G7, G8 and G25 (Fig. 3e) and under irrigated condition were G6, G25, G2 and G24 (1F). Across the years under rainfed condition, G6, G19, G20 and G2 were selected (Fig. 3g) and under irrigated condition G20, G6, G24 and G18 were selected.
Gain (%) for each studied traits in 25 durum wheat genotypes under rainfed and irrigated conditions for each and across years. DHE: days to heading; DMA: days to maturity; PLH: plant height; TKW: 1000-kernel weight; YLD: grain yield.
The SD for phenological traits was negative for all environments, showing enhanced earliness in selected genotypes. For grain yield positive SD for all environments was observed, that ranged from 1.63% in irrigated environment in 2021–2022 to 16.4% in irrigated environment in 2019–2020 (Fig. 5, Table 3). For TKW-positive gains varied between 2.21% (rainfed environment in 2020–2021) and 9.15% (rainfed environment in 2021–2022) and for plant height positive gains ranged from 2.93% (rainfed environment in 2019–2020) to 12.1% (rainfed environment in 2020–2021). These results show that the selected genotypes stand out as having high mean performance with better agronomic characteristics in contrasting environments.
The selected breeding lines expressed high values for positive gains (high performance, kernel weight and plant stature) and low values for negative gains (phenological traits), which indicate their relevance in durum breeding programme for desired traits. For example, across rainfed conditions, selected genotypes (breeding lines G6, G19, G20 and G2) combined high plant height, high grain yield, high 1000-kernel weight and low days to maturity (early maturity) and heading and under irrigated condition selected genotypes (breeding lines G20, G6, G2, and G18) combined low days to heading, high days to maturity, high grain yield, high 1000-kernel weight and high plant height. These genotypes are highlighted as candidates for cultivar recommendations as well as potential generators to obtain segregating populations with desired traits.
Factor analysis for drought tolerance indices
Table 4 presents the results of factor analysis for 25 durum wheat genotypes based on eight drought tolerance indices in each and across cropping seasons. Stress intensity (SI) in 2019–2020, 2020–2021, 2021–2022 and across years were 0.05, 0.35, 0.31 and 0.20, showing different drought stress from mild to relatively severe. The low SI in 2019–2020 was due to high rainfall received during early heading under rainfed condition, which significantly reduced the terminal drought stress. This situation provided no significant difference between rainfed and irrigated conditions in 2019–2020.
Eigenvalues, explained variance, factorial loadings after varimax rotation and communalities obtained in the factor analysis for 25 durum wheat genotypes based on eight drought tolerance indices in single and across cropping seasons. The bold values represent the indices with high contribution to each factor
FA1, FA2 stand for the first and second factors, respectively; The bold values represent the traits with high contribution to each factor. SI: stress intensity; Ys: yield under rainfed condition; Yp: yield under irrigated condition; MP: mean productivity; GMP: geometric mean productivity; STI: stress tolerance index; YI: yield index; YSI: yield stability index; DRI: drought response index; TOL: tolerance; SSI: stress susceptibility index.
In 2019–2020, the first two factors with eigen values >1 were retained, and totally accounted for 99.6% of the variance, with average communality equal to 0.975. The first factor closely correlated with indices of STI, GMP, MP, YI, Ys and Yp, while the second factor closely associated with YSI, TOL, SSI and DRI. In 2020–2021, the first two factors captured 95.5% of total variation, with the average communality of 0.995. The first factor represented the highest contribution for the indices STI, GMP, MP, YI, Ys and Yp, whereas second factor represented the highest contribution for YSI, SSI, TOL and DRI. The results of factor analysis for drought tolerance indices in 2021–2022 showed that the first two factors captured for 99.6% of variation. The indices STI, GMP, MP, YI, Ys,and Yp closely associated with first factor, while the indices YSI, SSI, TOL and DRI contributed most to second factor. Across years, the first two factors accounted for 97.2% of total variation. The first factor associated with MP, STI, GMP, YI, Ys,and Yp and second factor was correlated with SSI, TOL, YSI,and DRI (Table 4).
Genotypes selection for drought tolerance in variable stress conditions
The genotypes selected based on drought tolerance indices by the MTSI under mild drought stress condition (2019–2020 cropping season) were breeding lines G7, G8, G16 and G10 (Fig. 6a). Breeding lines G13 and G22 also were close to the cut point (red line that indicates the number of genotypes selected according to the selection pressure), which suggests that these two genotypes also can present interesting features for drought tolerance. The strengths and weaknesses view of the genotypes (Fig. 7a) indicated that the first factor exhibited the highest contribution for genotype G16, followed by G10, while second factor represented the highest contribution for genotype G7, followed by G8. Breeding lines G16 and G10 had strengths related to first factor represented by high values of STI, MP, GMP, YI and mean yields under rainfed (Ys) and irrigated (Yp) conditions, while genotypes G7 and G8 most associated with second factor and thus are characterized as resistant genotypes based on SSI, YSI, TOL and DRI.
Genotype ranking in ascending order for the MTSI based on eight drought tolerance indices in single (2019–2020, 2020–2021 and 2021–2022) and across three years (2019–2022). The selected genotypes based on this index are shown in red. The central red circle represents the cut point according to the selection intensity of 15%.
The strengths and weaknesses view of the selected genotypes for eight investigated drought tolerance indices based on MTSI are shown as the proportion of each factor on the factor analysis in each rainfed and irrigated trials during three cropping seasons (2019–2022). The smallest the proportion explained by a factor (closer to the external edge), the closer the traits within that factor are to the ideotype. The dashed line shows the theoretical value if all the factors had contributed equally.
In the line of moderate drought stress condition in 2020–2021 (Fig. 6b), breeding lines G20, G5, G3 and G15 were identified as high-performing and drought-tolerant. Zahab cultivar (G2) also was near to the cut point, showing good potential for drought tolerance. Evaluation of genotypes through the strengths and weaknesses view of the genotypes (Fig. 7b) showed that the breeding lines G5 and G15 exhibited strengths related to first factor and can be characterized as genotypes with highest values of STI, GMP, MP, YI with high performance under both rainfed and irrigated environments. Breeding lines G3 and G20 with highest association with second factor and then characterized as resistant genotypes based on SSI, YSI, TO and DRI, being best performing in rainfed conditions with yield stability in both conditions.
In the case of moderate drought stress condition in 2021–2022, breeding lines G4, G5, G22 and G19 (Fig. 6c) were selected as most drought-tolerant genotypes. Breeding lines G6 and G3 were also close to the cut point, indicating their potential for drought tolerance. According to the strengths and weaknesses view of the genotypes (Fig. 7c), genotype G4 associated with the first factor and is characterized by high MP, GMP, STI, YI and high mean grain yield under both rainfed and irrigated environments; while breeding lines G22, G5 and G19 contributed most to second factor, characterized as resistant genotypes based on TOL, SSI, YSI and DRI, showing high response to drought with stable yield in contrasting conditions.
Across three years, breeding lines G20, G5, G16 and G7 (Fig. 6d) expressed highest performing and drought tolerance. Genotype G20 showed strengths related to the first factor and is characterized by high MP, STI, GMP, Yp, Ys and YI (Fig. 7d). Genotypes G16, G7 and G5 showed strengths related to the second factor and then characterized as resistant genotypes based on SSI, YSI, TOL and DRI.
Interrelationships between drought tolerance indices
A heat map-based correlation analysis was applied to study the relationships among drought tolerance criteria in each and across years (Fig. 8). The indices of STI, GMP and MP were closely correlated (P < 0.01) to each other and to mean yield in both rainfed and irrigated environments in each and across cropping seasons, suggesting usefulness of these indices for selection of high drought-tolerant genotypes in stress and nonstress conditions. Three indices of YI, DRI and YSI were significantly associated (P < 0.01) with each other and to grain yield only under rainfed condition, indicating that selection based on these three indices will increase genotype performance under drought condition. The TOL and SSI were closely correlated to each other and significantly and positively correlated with grain yield under irrigated conditions but negatively correlated under rainfed conditions.
Heat map showing correlation analysis between mean yields under both rainfed (Ys) and irrigated (Yp) conditions and eight drought tolerance indices for 25 durum wheat genotypes in 2019–2020 (a), 2020–2021 (b), 2021–2022 (c) and across three cropping seasons (d).
For better classification and separation of 25 different genotypes based on drought tolerance indices, a PCA-based biplot was constructed for each and across years (Fig. 9). The first two PCs explained 97.7% to 99.6%, of the total variation in dataset. The close correlation between STI, MP and GMP and significant positive correlation (P < 0.01) of these three indices with mean yield in the both conditions, indicated that these indices were able to discriminate G20 and G16 in 2019–2020 (Fig. 9a); G6 and G5 in 2020–2021 (Fig. 9b); G4 in 2020–2021 (Fig. 9c) as the drought-tolerant group with high mean performance in both conditions. The SSI and TOL were closely associated with genotypes ranking and were correlated with mean performance under irrigated condition. Genotypes G1, G6 and G24 in 2019–2020 (Fig. 9a); G1, G12, G11 and G10 in 2020–2021 (Fig. 9b); and G11 in 2021–2022 (Fig. 9c) exhibited susceptibility to drought stress, although they performed well under irrigated conditions. The DRI and YSI were closely associated with each other and were significantly and positively correlated with YI and mean yield under rainfed condition. Genotypes G13, G7 and G8 in 2019–2020 (Fig. 9a); G20, G3 and G16 in 2020–2021 (Fig. 9b); and G22, G7 and G5 in 2021–2022 (Fig. 9c) positively interacted with these indices and were considered as high responsive genotypes to rainfed condition and good stability in rainfed conditions.
PCA-based biplot analysis for mean yield of 25 durum wheat genotypes (G1-G25) under both rainfed (Ys) and irrigated (Yp) conditions and eight drought tolerance indices in 2019–2020 (a), 2020–2021 (b), 2021–2022 (c) and across three cropping seasons (d).
Taking into account the three cropping seasons, the first two PCs explained 99.41% of the total variation in matrix data (Fig. 9d). The three indices of STI, MP and GMP were closely correlated to each other and to mean performance under rainfed and irrigated conditions, indicating that selection based on these indices will discriminate genotypes with high tolerance to drought and performance in the both rainfed and irrigated conditions. These indices were able to discriminate breeding lines G20 and G5 as drought tolerant genotypes with high performance in the both conditions. The yield index (YI) showed close correlation with Ys and positive correlation with DRI, STI, MP and GMP. Based on this index, breeding lines G20, G5 and G16 were identified as best-performing genotypes under drought condition.
Based on positive correlation between DRI and YSI, breeding lines G7, G8, G16 and G22 seemingly exhibited positive response to drought and good stability across both conditions. The SSI and TOL were closely correlated with each other and showed conversely correlations with DRI and YSI. Genotypes G1 and G24 positively interacted with SSI and TOL and could be considered to be susceptible to drought and thus unsuitable in rainfed condition with high instability grain yield under both conditions. Genotypes G15, G10 and G18 were placed around the biplot origin, and thus expressed an intermediate rank for different indices, and so they could be grouped as semi-tolerant or semi-sensitive genotypes to drought stress.
Agronomic characteristics of selected drought-tolerant genotypes
Agronomic characteristics of four top drought-tolerant genotypes selected by multi-trait stability index plot in each and across years is presented in Table 5. The results showed that the four top selected drought-tolerant genotypes (G7, G18, G16 and G10) in 2019–2020 in rainfed condition had higher mean yields than grand mean. In the case of TKW, breeding lines G7, G18 and G16 expressed higher TKW than grand mean; while for plant height only G16 exhibited higher plant stature than grand mean. Under irrigated condition, G16 showed higher mean yield, TKW and plant stature than grand mean, while G10 showed high mean yield than grand mean.
Agronomic characteristic of four top genotypes selected by multi-trait stability index applied on drought tolerance indices across three cropping seasons in rainfed and irrigated environments
DHE: days to heading; DMA: days to maturity; PLH: plant height; TKW: 1000-kernel weight; YLD: grain yield.
In 2020–2021 under rainfed condition, all four selected drought-tolerant genotypes (G20, G3, G5 and G15) showed higher mean yields than the grand mean; G20 also expressed higher TKW and plant stature and earliness than grand mean. Under irrigated condition G5, G15 and G20 expressed higher mean yield and plant height than the grand mean, and G5 and G20 exhibited higher TKW than grand mean. The selected drought-tolerant genotypes exhibited lower or equal values of days to heading and days to maturity compared to the overall average. In 2021–2022 under rainfed condition, selected genotypes (G4, G5, G22 and G19) exhibited better performance than grand mean, while under irrigated conditions G4, G5 and G19 expressed higher mean yields than the grand mean. In the both conditions, G4 exhibited higher TKW than grand mean.
When considering the mean values of agronomic characteristics for the selected drought tolerant genotypes in each and across cropping seasons for all measured traits, their superiority for investigated traits is evidenced (Table 5). This results show that selected genotypes based on drought tolerance indices also exhibited good agronomic performance.
Discussion
In this study, we assessed agronomic performance and drought tolerance of 25 durum wheat genotypes based on five basic phenological and agronomic traits and several drought stress criteria. The results indicated a significant effect of GE interaction for all traits, proving that the trait performance of genotypes was strongly affected by different environmental conditions. Durum wheat is generally cultivated without supplemental irrigation under optimal rainfed conditions with rainfall around 450 mm and monthly normal distribution. Rainfall in cropping seasons 2020–2021 and 2021–2022 showed 30% and 50% reduction compared to normal condition, while it showed 15% increase in 2019–2020. The significant annual climate change was mainly due to unprecedented amount of rainfall particularly during stem elongation to grain filling in second and third cropping seasons, rather than the drought effect caused during the first cropping seasons. Increasing temperature in heading and grain filling period in second and third cropping seasons also result in an additional reduction in yield. In comparison to the second and third cropping seasons, the durum wheat genotypes benefited from perfect environmental conditions during the first growing season (2019–2020), as the supplemental irrigation in this cropping season only 5% increased yield, while in second and third seasons increased up to 35% and 31%, respectively, in grain yield compared to rainfed condition. Similarly, genetic diversity of drought tolerance of 25 durum wheat genotypes was studied based on drought tolerance criteria using multivariate technique. The results showed that there is a high genetic diversity in the drought tolerance of the genotypes, which can be investigated in the durum breeding program.
Factor analysis showed close correlation between grain yield, TKW and plant height and negative correlation of these traits with phenological traits in each and across environments, indicating that selection of genotypes based on grain yield results in choosing genotypes possessing high TKW and plant stature with earliness. This is in agreement with previous studies that indirect selection for phenological traits and kernel weight under rainfed conditions increases grain yield of durum wheat (Mohammadi et al. Reference Mohammadi, Armion, Sadeghzadeh, Amri and Nachit2011b; Mohammadi and Amri Reference Mohammadi and Amri2013). Therefore, selection based on multi-trait stability index may collect genotypes with superior adaptability to the prevailing climatic conditions, which is very important for hybridization programmes (Al-Ashkar et al. Reference Al-Ashkar, Sallam, Almutairi, Shady, Ibrahim and Alghamdi2023; Hussain et al., Reference Hussain, Akram, Shabbir, Manaf and Ahmed2021; Olivoto et al., Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a, Reference Olivoto, Lúcio, Silva, Sari and Diel2019b). According to MTSI genotypes G6, G19, G20 and G2 showed highest performing and stable across rainfed environments, while across irrigated environments breeding lines G20, G6, G24 and G18 expressed highest performance and stability based on the investigated traits. These genotypes show the greatest potential for the simultaneous improvement of the measured traits in durum wheat breeding programmes. The MTSI identified the two breeding lines G6 and G20 as common genotypes in both rainfed and irrigated conditions. These genotypes showed high trait stability and performance across the both conditions across three cropping seasons.
In this study, we used a heat-map correlation analysis and PCA-based biplot for grouping of genotypes and drought tolerance criteria (Figs. 8 and 9). Positive and highly significant correlations were found between mean grain yield with STI, GMP and MP in both rainfed and irrigated conditions, showing the efficiency of these indices for genetic improvement in drought tolerance and high performance in contrasting conditions. Findings on PCA-based biplot analysis revealed different sets of positive interactions between indices and genotypes that can be useful in selecting parents for a breeding programme. In addition, the correlation pattern between the indices shows that the selection based on these indices has the potential to identify superior lines in durum breeding programme. However, the selected genotypes by this approach were in accordance with those selected by MTSI plot (Fig. 6).
The results indicated that using the MTSI technique developed by Olivoto et al. (Reference Olivoto, Lúcio, Silva, Marchioro, Souza and Jost2019a) to select genotypes with traits stability performance, successfully helped to identify genotypes with high and stable performance and the expected gains from selecting those genotypes for the investigated traits. However, the potential of this approach for the simultaneous improvement of the multiple traits using predicted genetic effects has been reported in previous studies (Norman et al., Reference Norman, Agre, Asiedu and Asfaw2022; Olivoto and Nardino, Reference Olivoto and Nardino2021; Sellami et al., Reference Sellami, Pulvento and Lavini2021; Zuffo et al., Reference Zuffo, Steiner, Aguilera, Teodoro, Teodoro and Busch2020; Yue et al., Reference Yue, Olivoto, Bu, Li, Wei, Xie, Chen, Peng, Nardino and Jiang2022).
The pattern view of strengths and weaknesses of genotypes, which shows the ratio of multi-trait stability index explained by each factor, is an important tool to identify the strengths and weaknesses of studied genotypes (Al-Ashkar et al., Reference Al-Ashkar, Sallam, Almutairi, Shady, Ibrahim and Alghamdi2023; Norman et al., Reference Norman, Agre, Asiedu and Asfaw2022). Based on our findings on drought stress condition, genotypes varied but some sharing genotypes were detected across environments. Knowledge of the contribution of these genotypes helps to select possible genitors for future crosses. Across three years, breeding lines G20, G5, G16 and G7 (Fig. 3d) expressed highest performing and drought tolerance. Genotype G20 exhibited strengths for high MP, STI, GMP, Yp, Ys and YI, while G16, G7 and G5 showed strengths for high YSI and DRI and low values of SSI and TOL, suggesting best performance of these genotypes under rainfed conditions. The genotype G20 showed best expression based on drought stress indices and agronomic traits. Thus, this genotype can be recommended as new genetic resources for improving and stabilizing the grain yield in durum wheat programmes under drought and irrigated conditions. Hence, these methods, if jointly used can serve as a powerful tool to assist breeders in MET.
Based on the heat map correlation matrix (Fig. 8) and PC (Fig. 9) analyses, the STI, GMP and MP were closely correlated (P < 0.0) to each other and to mean yield in both rainfed and irrigated environments in each and across cropping seasons, suggesting the efficiency and usefulness of these indices for selection of high drought-tolerant genotypes in both conditions. Findings are in agreement with Mohammadi (Reference Mohammadi2016), Ayed et al. (Reference Ayed, Othmani, Bouhaouel and Teixeira da Silva2021), and Nouri et al. (Reference Nouri, Etminan, Teixeira da Silva and Mohammadi2011), who reported that these indices are mostly efficient and useful to differentiate durum wheat genotypes. In the present work, STI, MP and GMP identified G20 and G16 in 2019–2020, G6 and G5 in 2020–2021 and G4 in 2020–2021 as the most drought-tolerant genotypes with high mean performance under drought irrigated conditions. These genotypes have a good tolerance compared to drought-tolerant check genotypes (G1 and G2). Therefore, this group of genotypes were well adapted in a specific environment under drought conditions with irregular distribution of rainfall, insufficient rainfall and extreme temperature events.
Conclusion
The assessment of drought tolerance in both drought and irrigated environments across three years led us to conclude that MTSI could be used to select promising durum wheat lines with improved yield traits and drought tolerance. The investigated 25 durum wheat genotypes were categorized into groups based on their performance under drought and irrigated environments. We were able to identify genotypes that showed differential response under drought and irrigated environments while some performed well under both set of environments. Based on traits studied, MTSI identified breeding lines G20, G6, G25 and G18 as high-yielding genotypes with traits stability. Based on drought tolerance indices, the MTSI identified breeding lines G20, G16, G7 and G5 as stable drought tolerant across variable stress conditions. Breeding line G20 showed highest trait stability performance and drought tolerance across environments. In our study, the MTSI was a useful tool for selecting genotypes based on their agronomic performance and drought tolerance that could be exploited for identification and selection of elite genotypes with desired multi-traits. In conclusion, this study resulted in identifying several outstanding breeding lines that out-yielded the check cultivars based on agronomic performance and drought tolerance which should be recommended directly for cultivar recommendation or to be crossed with commercial cultivars for improve agronomic performance and drought tolerance in durum wheat under the new circumstances of climatic change.
Acknowledgements
We thank the two reviewers and the editor of Experimental Agriculture for providing helpful comments and corrections to the manuscript.
Financial support
This research was conducted based on a part of the project number 0-15-15-062-981039, founded by the Dryland Agriculture Research Institute (DARI) of Iran.
Competing interests
The authors declare no conflicts of interest.
