Do modern winter wheat cultivars outperform traditional landraces under cold stress? A large-scale agro-morphological evaluation

Abstract

Background

Cold stress remains a primary abiotic constraint limiting wheat (Triticum aestivum L.) productivity in high-altitude environments. While landraces represent an underexplored reservoir of adaptive traits, integrating their genetic diversity with modern breeding goals requires advanced selection tools. This study evaluated the agro-morphological performance, quality characteristics, and cold hardiness of 180 purified wheat landraces together with six registered cultivars (checks), totaling 186 genotypes from Eastern Anatolia. Using an augmented experimental design over two growing seasons in Erzurum, Türkiye, we aimed to integrate stability parameters with modern multi-trait selection indices (MGIDI and FAI-BLUP) to identify superior genotypes for continental climates. Adjusted means were obtained using a mixed model approach, and genotype effects were estimated as best linear unbiased predictions (BLUPs), which were subsequently used for multivariate analyses.

Results

Substantial genotypic variation was observed across all traits. A significant yield gap was detected, with registered cultivars outperforming landraces in mean grain yield (2751.9 kg ha⁻¹ vs. 1556.5 kg ha⁻¹), whereas landraces demonstrated superior grain physical properties, particularly in thousand-kernel weight. Cold tolerance screening identified critical survival thresholds: all cultivars and 28 landraces survived at − 13 °C, while only seven landraces-maintained viability at − 15 °C. At − 19 °C, only the cultivar Alparslan survived. MGIDI analysis confirmed the agronomic superiority of modern cultivars (G181–G186), which ranked closest to the multi-trait ideotype at a 15% selection intensity. However, specific landraces (G81 and G173) emerged as highly divergent genotypes, successfully bridging the gap between landrace-driven grain quality and cultivar-driven cold resilience.

Conclusion

The integration of multi-trait indices enabled more robust identification of elite germplasm compared to traditional ranking approaches. While modern cultivars demonstrated a clear adaptive advantage in cold-prone environments, the identified elite landraces serve as valuable reservoirs of adaptive genetic variation. Genotypes G183, G181 (Alparslan), G81, and G184 are recommended as primary parental resources for breeding programs aiming to combine high yield potential with extreme cold resilience under fluctuating climatic conditions. This study offers one of the first large-scale integrations of traditional landrace diversity with modern multi-trait selection indices under extreme continental conditions, providing a practical framework for ideotype-oriented breeding in cold-prone environments.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09431-8.

Introduction

Wheat (Triticum aestivum L.) is one of the most widely cultivated cereal crops worldwide and constitutes a primary staple food source for nearly 35% of the global population [1]. Global wheat production reached approximately 778.64 million metric tons during the 2021–2022 growing season; however, current production levels are considered insufficient to meet the increasing demand of a global population projected to reach nearly 10 billion by 2050 [2]. It is estimated that global cereal production must increase by approximately 70% to ensure future food security [3]. Consequently, enhancing the adaptive capacity of wheat to environmental stresses has become a central objective in contemporary breeding and genetic improvement programs. Abiotic stresses, including drought, salinity, extreme temperatures, and oxidative stress, represent major constraints on agricultural productivity, leading to yield losses that may exceed 50% of potential production annually [4, 5]. Among these factors, cold stress encompassing both chilling (0–15 °C) and freezing (< 0 °C) temperatures stands out as a critical limitation to wheat productivity, particularly in temperate agro ecosystems (Ahad et al., 2023; Hassan et al., 2021). Low temperature conditions disrupt essential physiological and biochemical processes, including membrane stability, protein folding, enzymatic activity, and photosynthesis, ultimately resulting in cellular dehydration and reduced plant growth [6, 7]. Therefore, a comprehensive understanding of cold stress tolerance mechanisms is essential for improving wheat adaptation and stability under fluctuating environmental conditions.

Wheat cultivars are classified based on their vernalization requirements into winter, semi winter, weak spring, and spring types. Winter wheat requires exposure to a defined period of low temperatures to complete the vernalization process and initiate reproductive development [8]. Although winter cultivars exhibit greater tolerance to low temperatures compared to spring types, exposure to severe freezing conditions can cause substantial cellular damage. Temperatures below 0 °C disrupt membrane integrity, alter cytoplasmic structure, and may ultimately lead to plant death [9]. Previous studies indicate that spring wheat generally cannot survive temperatures around − 10 °C, whereas winter wheat genotypes may tolerate temperatures below − 15 °C depending on their genetic background (Ejaz et al., 2023). This capacity, defined as freezing tolerance, is a complex quantitative trait governed by multiple genes and associated physiological and biochemical processes [5, 10]. Accordingly, elucidating both genetic and phenotypic components of freezing tolerance remains a key priority in wheat genetics and breeding research. In winter wheat, freezing tolerance is acquired through the process of cold acclimation, which occurs under low but nonfreezing temperatures during autumn. However, climate change has increased the frequency of erratic temperature fluctuations, thereby reducing the efficiency of acclimation processes. In particular, the reduction of snow cover normally serving as a protective insulation layer has significantly increased the risk of winterkill and spring frost damage [11]. It is estimated that approximately 85% of global wheat growing areas are affected by spring frost events during critical developmental stages, resulting in yield losses of up to 90% under severe conditions [12]. Enhancing cold and winter resistance is therefore a key strategy to mitigate yield losses and improve crop resilience under climate variability [13]. The impact of climate change on winter crop survival is particularly pronounced in regions characterized by high climatic variability. Warmer autumn periods followed by sudden freezing events significantly reduce overwinter survival rates [14, 15]. Moreover, reduced snow cover exposes plants to extreme temperature fluctuations, further increasing the risk of cold injury and yield loss [11, 15]. These combined stress factors underscore the necessity for region specific and genetically informed strategies to improve cold stress resilience in wheat.

In Türkiye, wheat production holds substantial socio-economic importance; however, it is strongly influenced by the country’s diverse and often unpredictable climatic conditions. Major wheat producing regions, including Central, Eastern, and Northeastern Anatolia, are frequently exposed to cold stress during winter and early spring periods. Historical evidence indicates that favorable climatic conditions may support high yield levels, whereas sudden cold waves and spring frosts can lead to severe productivity losses [13]. Furthermore, although modern high yielding cultivars have significantly improved productivity, they often lack sufficient stability under abiotic stress conditions, leading to yield fluctuations. This situation highlights the urgent need for the development of landrace adapted and stress-resilient wheat cultivars. Developing cultivars with stable yield performance under fluctuating environmental conditions remains a primary objective of modern wheat breeding. Success in breeding programs targeting quantitative traits in various crop plants is highly dependent on the presence of genotype-by-environment (G × E) interactions. The more severe this interaction, the more difficult it becomes to select desirable genotypes based solely on their phenotype. The G × E interaction is of distinct importance for complex quantitative traits like grain yield, which are controlled by a vast number of genes. Generally, genotypes that exhibit superior performance across diverse environments compared to others are considered ideal genotypes [16]. However, progress in conventional breeding is increasingly constrained by the narrow genetic base of elite germplasm, limiting the potential for improving complex traits such as cold tolerance [17]. Therefore, broadening the genetic base through the integration of diverse germplasm resources is essential for sustainable genetic gain. In this context, wheat landraces represent indispensable reservoirs of genetic diversity [18]. Having evolved under long term natural and farmer mediated selection, landraces exhibit high adaptability to marginal environments and enhanced tolerance to abiotic stresses [19]. These genotypes harbor valuable alleles that are often absent in modern cultivars. However, effective utilization of landraces requires precise characterization of their phenotypic and genetic diversity to identify divergent and superior parental lines [20, 21]. Thus, systematic evaluation of landrace germplasm is a prerequisite for their efficient incorporation into breeding programs. However, despite the known resilience of Anatolian landraces, there is a lack of large-scale studies integrating modern multi-trait selection indices like MGIDI and FAI-BLUP to evaluate their performance under the extreme continental climate of Eastern Anatolia.

Phenotypic characterization based on agro morphological traits remains a fundamental and practical approach in plant breeding despite advances in molecular techniques [22]. Since phenotype represents the integrated expression of genotype and environment, morphological evaluation provides critical insights into genotype performance under stress conditions. While qualitative traits are generally controlled by simple genetic mechanisms, quantitative traits such as yield under cold stress are polygenic and strongly influenced by environmental factors, making their selection more complex [23]. Considering other quantitative traits such as thousand-kernel weight, plant height, and spike length, as well as phenological characteristics like earliness and grain filling duration, is another vital step in breeding programs. Considering these traits has a significant impact on the selection and introduction of adaptable and stable cultivars suited for specific regions or a broad spectrum of target environments [24]. Given the increasing climatic variability, developing cold-resilient wheat landraces with superior agro-morphological performance has become a key breeding priority. However, evaluating multiple, often correlated traits can lead to multicollinearity issues, complicating selection decisions. Multi-trait selection approaches, such as the Multi-trait Genotype–Ideotype Distance Index (MGIDI), can overcome these challenges by integrating trait information through factor analysis and ranking genotypes based on overall performance [25, 26]. Therefore, integrative approaches combining phenotypic variation and multivariate statistical analyses are essential for effective genotype discrimination. However, traditional multivariate techniques such as PCA and cluster analysis do not provide a direct selection ranking, necessitating the use of ideotype-based multi-trait indices for balanced genotype selection. In this context, evaluating the extent to which agro morphological variation reflects cold stress tolerance among wheat landraces represents a critical step toward identifying superior genetic resources. Despite the recognized importance of wheat landraces as reservoirs of adaptive genetic variation, large-scale studies integrating advanced multi-trait selection indices such as MGIDI and BLUP-based ranking under extreme continental climates remain scarce. In particular, the combined evaluation of cold tolerance, yield components, and grain quality within a unified selection framework has not been sufficiently addressed in Eastern Anatolian germplasm. To the best of our knowledge, no previous study has simultaneously integrated large-scale Anatolian winter wheat landrace diversity, controlled freezing assays, BLUP-based genetic merit estimation, and MGIDI-driven ideotype selection under the extreme continental conditions of Eastern Anatolia. We hypothesized that modern cultivars outperform landraces in terms of yield stability and cold tolerance under extreme freezing conditions, while selected landraces retain unique adaptive traits related to grain quality and sink capacity, and the integration of BLUP-based estimates with MGIDI would significantly improve the accuracy of multi-trait genotype selection under strong genotype × environment interactions. Therefore, the objective of this study was to (i) quantify agro-morphological, quality, and cold tolerance variation among winter wheat genotypes, (ii) dissect their multivariate relationships using PCA and clustering, and (iii) identify superior ideotype-based genotypes using BLUP and MGIDI under high-altitude continental conditions.

Materials and methods

Plant materials

A total of 180 wheat landrace genotypes were used as experimental material in this study. These genotypes were originally collected from different regions of Türkiye within the framework of the International Winter Wheat Improvement Program (IWWIP). Following collection, the landraces were purified through single spike selection, resulting in 42 genotypes from Bitlis, 3 from Van, 113 from Ağrı, and 22 from Bayburt. In addition to the landraces, six registered wheat cultivars were included as reference genotypes in the study. Thus, the experimental material consisted of 180 landrace-derived lines and six commercial cultivars (checks), totaling 186 genotypes. Doğu 88, Palandöken 97, Alparslan, and Ayyıldız, developed by the Eastern Anatolia Agricultural Research Institute, are winter type cultivars adapted to rainfed conditions and characterized by tolerance to cold and drought stress. These cultivars differ in maturity, grain quality, and disease resistance levels. Gerek 79, released by the Transitional Zone Agricultural Research Institute, is a drought adapted cultivar with moderate yield potential but relatively higher susceptibility to rust diseases. Karahan 99, developed by the Bahri Dağdaş International Agricultural Research Institute, is a winter type cultivar with good adaptability to Central Anatolian conditions and exhibits tolerance to cold, drought, and lodging. Detailed information regarding the landrace genotypes is presented in Table S1, while a summary of the characteristics of the registered cultivars used in the experiment is provided below.

Field implementation and crop management

Field trials were conducted during the 2013–2014 and 2014–2015 growing seasons at the Erzurum Ilıca experimental site (39° 57’ N, 41° 05’ E, altitude 1850 m), located in Eastern Anatolia. The climatic conditions during these periods showed significant variation, with mean temperatures ranging from − 13.9 °C in December 2014 to 22.2 °C in August 2014, reflecting the typical harsh winter and warm summer profile of the high-altitude Erzurum plateau (Table 1 and Fig. 1). The 2014–2015 growing season exhibited a colder winter period compared to the previous year, which is a critical factor for evaluating the winter hardiness of the genotypes. The soil at the experimental site is characterized as a clay-loam texture with an organic matter content of 1.8%, a pH of 7.6, and low phosphorus availability. Winter sowing was performed within the optimal regional planting window (1 September–1 October) following established agronomic recommendations [27]. Seeds were mechanically sown at a depth of 4–6 cm, with 20 cm inter row spacing and a density of 475 seeds m⁻² [28]. An augmented block design was employed to accommodate the large number of genotypes while maintaining statistical reliability [29]. Adjusted means for unreplicated landrace entries were estimated using BLUPs derived from the mixed-model framework, where replicated check cultivars provided the basis for correcting environmental heterogeneity across blocks. The experimental design consisted of multiple augmented designs. In the augmented design, check cultivars were replicated across blocks, whereas test genotypes were unreplicated, allowing adjustment of genotype means through mixed-model analysis to account for environmental heterogeneity. Experimental plots were arranged with dimensions of 6 m in length and 1.2 m in width (Total area 7.2 m², with a net harvest area of 5.0 m² after removing border effects). Fertilization consisted of diammonium phosphate and ammonium nitrate (26% N). Phosphorus was fully applied at sowing, while nitrogen was split between sowing and stem elongation stages, with a total application rate equivalent to 60 kg N ha^{− 1} and 60 kg P₂O₅ ha^{− 1} [27].

Table 1.

Monthly average temperatures and total precipitation for the growing seasons at Erzurum

Year	2013–2014	2014–2015
Month	Mean Temperature (°C)
September	13.6	4.28
October	6.0	9.64
November	2.3	2.26
December	-9.7	-13.9
January	-6.4	-9.85
February	-4.5	-6.89
March	2.7	2.41
April	7.8	7.59
May	11.7	5.85
June	15.9	11.2
July	21.2	19.12
August	22.2	20.02

Fig. 1.

Average monthly temperature and precipitation of the experimental area during (a) 2013 and (b) 2015 years

Determination of morphological and yield traits

Data on various agro morphological and yield related traits were collected as follows: grain yield (kg ha^–1): after the cleaning of harvested wheat, grains weight was recorded, and yield was calculated accordingly. Grain yield values were converted to kg ha⁻¹ based on net plot area. Number of ears per m²: during the ripening period, the number of ears in two randomly selected rows (each 1 m long) from each plot was counted, and the values were extrapolated to represent the number of ears. Plant height (cm): at maturity, the length of the stem from 10 randomly selected spikes in the harvest area of each plot was from the soil surface to the top spikes of the ear. Heading time: the number of days from 1 January until the flag leaf emerged from the sheath in half of the ears in each plot was recorded as the heading time. Maturation time: the number of days between the heading date of the plants and the date they reached harvest maturity was recorded as the maturation time. Number of grains per spikes: 10 spikes were randomly selected from each plot, threshed separately, and the number of grains per spikes was recorded. Grain weight per spike (g): the seeds obtained from 10 spikes were weighed separately using a precision scale (0.001 g), and the average grain weight per spike was determined [13].

Quality parameters

Physicochemical traits, including 1000 grain weight, hectoliter weight, protein ratio (via NIR spectroscopy), Zeleny sedimentation, and gluten characteristics (wet/dry gluten and gluten index), were quantified following ICC standard protocols No. 116/1 and 158 [30]. For 1000 grain weight (g): four samples, each containing 100 grains, were selected and weighed from each plot, and the average weight was calculated. The 1000 grain weight was determined by multiplying the average weight by 10. Hectolitre (hL) weight: three random samples of wheat seeds were collected from each plot, weighed using a hL apparatus, and the average value was calculated. Protein ratio (%): a 100 g wheat sample from each plot was ground using a crushing mill, and the protein content was determined via near infrared (NIR) spectroscopy, using a 20 g flour sample. Zeleny sedimentation (mL): determined by taking a 3.2 g flour sample from the wheat harvested from each plot, grinding it in a crushing mill, and adding bromine phenol blue solution and lactic acid, in accordance with ICC Standard No. 116/1 [31]. The methodology followed the protocol of P Williams, MG Lindhauer, RE Poms, RL Wehling, W Bergthaller and CS Gaines [32]. Wet gluten (%): this refers to the elastic substance formed by the swelling of gliadin and glutenin proteins in wheat when they absorb water. Wet gluten content and gluten index in flour samples were determined according to ICC Standard No. 158 [31]. Gluten index (%): the wet gluten was centrifuged using a Glutomatic system, and the gluten index value was calculated as outlined by M Düzgün and M Karaman [33]. Dry gluten (%): dry gluten was determined by drying the wet gluten in the Glutomatic system for 5 min, followed by cooling in a desiccant container and weighing the resulting gluten mass [34].

Controlled cold stress assessment

Cold tolerance was assessed under controlled environmental conditions using L Gusta and D Fowler [35] protocol. Seeds were first stratified at + 4 °C in darkness for 48 h for break dormancy. Subsequently, they were transferred to room temperature (20–22 °C) and incubated for 24 h to promote germination. Germinated seeds were transferred into trays (40 × 25 × 7 cm) containing a homogenized substrate mixture (1:1:1 ratio of sand, field soil, and peat). Ten seedlings per genotype were grown under controlled conditions with three replicates. Plants were maintained for 49 days under a light intensity of approximately 300 µmol m⁻² s⁻¹ for cold acclimation. The 49-day acclimation period was selected to ensure sufficient cold hardening prior to freezing treatment, consistent with established protocols for winter wheat cold tolerance evaluation. Cold acclimation was performed under controlled photoperiod and temperature conditions prior to freezing treatment. Following this growth period, plants were subjected to progressive freezing treatments. The temperature gradually reduced from 0 °C to − 4 °C at a rate of 1 °C per hour and maintained at − 4 °C for 4 days. Subsequently, the temperature was lowered to − 7 °C and then decreased by 2 °C per hour until reaching − 13 °C, − 15 °C, − 17 °C, and − 19 °C. Independent sets of plants were removed at each temperature threshold. Post freezing recovery was carried out at 16 °C under a 16 h light / 8 h dark photoperiod for 3 weeks. Survival was assessed based on plant viability, and genotypes were classified according to their ability to maintain at least 50% survival under each temperature level. Survival percentage was calculated as the proportion of living plants relative to the total number of plants per genotype.

Statistical and multivariate analysis

All recorded traits were subjected to descriptive statistics and analysis of variance (ANOVA) using JMP software to determine the significance of genotypic differences. A mixed linear model was adopted to estimate genotype effects by considering genotype, year, and genotype × year interaction effects as random factors, allowing the prediction of genotypic performance across heterogeneous environments. Prior to analysis, assumptions of normality and homogeneity of variances were verified.

Pearson’s correlation coefficients were calculated to quantify relationships among traits, and the correlation matrix was visualized graphically. To obtain more reliable estimates of genotypic performance, best linear unbiased predictions (BLUPs) were computed under a mixed-model framework considering genotypes as random effects [36]. The mixed model framework enabled the separation of genetic effects from environmental noise, providing unbiased estimates of genotypic performance under heterogeneous field conditions. The mixed model used for BLUP estimation was defined as: Yijkl = µ + Gi + Yj + Bk(j) + GYij + εijkl where µ is the overall mean, Gi is the random genotype effect, Yj is the random year effect, Bk(j) is the random block effect nested within year, GYij represents the genotype × year interaction, and εijkl is the residual error. In the BLUP framework, genotype, year, block, and genotype × year interaction effects were treated as random to improve the accuracy of genetic merit prediction under variable environmental conditions. The genetic correlation (rG) between the two growing seasons was evaluated to ensure the consistency of the estimated breeding values. The standardized BLUP values were subsequently used to generate a heatmap to visualize genotype–trait associations and phenotypic variation patterns. Prior to multivariate analysis, all traits were rescaled and standardized to eliminate scale differences while preserving the correlation structure among variables. Multivariate relationships among genotypes were further explored using principal component analysis (PCA), and genotype grouping was performed through agglomerative hierarchical clustering based on Euclidean distance. In addition, the multi-trait genotype ideotype distance index (MGIDI) was calculated to identify superior genotypes by simultaneously considering all measured traits and their desired selection direction [26]. The ideotype was defined as a hypothetical genotype combining maximum yield, superior grain quality, and minimum cold damage, representing the optimal breeding target under continental stress conditions. Prior to the MGIDI analysis, all variables were checked for normality using the Shapiro-Wilk test, and logarithmic transformations were applied to traits with high skewness to fulfill the assumptions of factor analysis.

All multivariate and graphical analyses were conducted in R software. Pearson correlation analysis was performed using the cor () function and visualized with the corrplot package. Principal component analysis was carried out using the PCA () function from the FactoMineR package, and PCA plots were generated using factoextra [37]. Hierarchical clustering was performed using dist () and hclust () functions. Heatmap visualization was generated using the ComplexHeatmap package based on standardized BLUP values. BLUP estimates and multi-trait genotype selection were conducted using the metan package [26], and the MGIDI index was computed using the mgidi () function. These approaches allowed a comprehensive evaluation of phenotypic diversity, genotype relationships, and multi-trait superiority under cold stress conditions.

Factor analysis was performed to explore the multivariate structure of the measured traits and to identify the major latent factors explaining the observed variation among genotypes. Prior to analysis, all variables were standardized to zero mean and unit variance to eliminate scale effects. The number of retained factors was determined according to the Kaiser criterion (eigenvalue > 1), and factor extraction was conducted using principal component-based factor analysis. The optimal number of factors was determined using parallel analysis. Factor extraction was conducted using the maximum likelihood (ML) method, followed by varimax rotation to improve interpretability of factor structure. Factor loadings, communalities, unique values, and factor scores were obtained for each trait and genotype, respectively. In addition, genotypes were ranked within each factor according to their factor scores to identify superior entries associated with specific trait complexes. All analyses were performed in R using the psych package. Sampling adequacy for factor analysis was additionally evaluated using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s test of sphericity to confirm the suitability of the correlation matrix for dimensional reduction. To identify the most cold-tolerant and high-yielding genotypes among the bread wheat populations evaluated in this study, the Multi-Trait Genotype–Ideotype Distance Index (MGIDI) was employed. This multivariate selection approach enables the simultaneous assessment of multiple traits by summarizing the multi-trait variation into a single distance-based index, thereby facilitating the identification of genotypes closest to the ideotype. Prior to MGIDI computation, trait values were rescaled to a 0–100 range according to the desired selection direction (increase or decrease), preserving the original correlation structure among traits. The MGIDI analysis was performed in the RStudio environment using the metan package (version 1.18.0 or later) in R (version 4.5.1). The index was calculated according to the Euclidean distance between each genotype and the ideotype in the space factor, as follows:

where denotes the MGIDI value of the genotype, is the score of the genotype for the factor, , and represents the score of the ideotype for the corresponding factor. Genotypes with lower MGIDI values were considered closer to the ideotype and therefore more desirable.

A selection intensity of 15% was adopted to identify the superior genotypes. In defining the ideotype, the direction of selection was specified according to breeding objectives. An increase was desired for yield-related traits, such as grain yield, thousand-grain weight, and grain weight, whereas a decrease was targeted for cold damage scores (C1–C4) and phenological traits such as days to emergence and days to heading, to select earlier and more cold-tolerant genotypes.

In addition, the strengths and weaknesses of the selected genotypes were investigated by estimating the contribution of each trait to the MGIDI index , as expressed below:

where is the squared distance between the genotype and the ideotype for the trait. Lower contribution values indicate that the genotype is closer to the ideotype for that particular trait, whereas higher values reveal traits that contribute more strongly to the genotype’s divergence from the ideotype.

The results were visualized using genotype ranking plots and strengths–weaknesses graphs generated with the metan and ggplot2 packages. Selection gains for each trait were also computed based on BLUP estimates to quantify genetic progress achieved through MGIDI-based selection. Although genotype × year interactions were evaluated using mixed-model approaches, advanced multi-environment stability analyses such as AMMI or GGE biplot were not performed because the experiments were conducted under a single-location framework across two growing seasons.

Results

Morphological variation among genotypes

Considerable morphological variation was observed among the evaluated wheat genotypes with respect to spike color, awn formation, spike pubescence, and grain color (Table S2, Fig. 2). Of the total experimental material, 12 genotypes belonged to club wheat, whereas the remaining entries were bread wheat. Spike color distribution showed substantial diversity, with 104 genotypes having red spikes, 70 having white spikes, and 12 having black spikes. All registered cultivars were awned, whereas among the landraces 110 were awned, 15 were tip awned, and 55 were awnless. Spike pubescence also varied markedly, with 96 genotypes exhibiting pubescent spikes and 90 genotypes glabrous spikes. Grain color assessment indicated that 131 genotypes were white grained and 55 were red grained. This broad range of morphological descriptors reflects the high phenotypic diversity preserved within the landrace wheat germplasm. Such variation is particularly important in landrace-based studies because these visible characters often reflect long term adaptation to contrasting agro ecological conditions and farmer selection practices. The presence of both bread and club wheat forms, together with marked variation in spike architecture and grain coloration, further supports the view that the studied material represents a broad and valuable genetic base for selection and breeding.

Fig. 2.

Visual representation of (a) spike morphology of various landrace genotypes and (b) the overall layout of the experimental area

Agronomic traits

Analysis of variance revealed highly significant differences among genotypes for heading time in both years and in the combined analysis (P < 0.01; Table 2). Year effects were also significant (P < 0.05), whereas the genotype × year interaction was highly significant (P < 0.01). Mean heading time was 165.30 days in 2013 and 161.75 days in 2014 (Table 2). Among the registered cultivars, Alparslan was the earliest heading genotype in both years (157.20 and 161.00 days, respectively), whereas Ayyıldız was the latest (163.20 and 168.00 days, respectively). Across years, the earliest landrace was genotype 67 (157.78 days), while genotype 50 was the latest (171.78 days). Nearly half of the landrace genotypes headed later than the registered cultivars. Furthermore, the number of days to heading showed highly significant genotypic differences (P < 0.01; Table 2), with the overall mean decreasing from 165.29 days in the first year to 161.73 days in the second year (avg 163.32 days; Fig. 3b). The range varied between 156 and 177 days in the first year and 155–169 days in the second year. The higher F-values in the combined analysis (F = 67.30) compared to individual years (5.47 and 17.96) and the low coefficient of variation (CV = 1.80–2.70%) with a combined LSD value of 1.64 demonstrated consistent genotypic effects.

Table 2.

Descriptive statistics and Analysis of Variance (ANOVA) for yield and yield-related traits of the evaluated genotypes across two consecutive growing seasons

Trait	Statistic	Year 1	Year 2	Combined (Year 1 and Year 2)
Plant height (cm)	Min	45.00	65.00	55.00
	Max	95.00	120.00	107.50
	Mean	70.00	92.50	84.72
	F-value	5.18**	7.87**	9.8675**
	LSD	15.49	12.86	14.05
	CV (%)	7.00	4.00	5.00
Number of days to heading	Min	156.00	155.00	156.50
	Max	177.00	169.00	172.00
	Mean	165.29	161.73	163.32
	F-value	5.47**	17.96**	67.30**
	LSD	6.06	2.34	1.64
	CV (%)	2.70	1.80	2.50
Number of spikes per m2	Min	76.00	88.00	82.00
	Max	480.00	528.00	490.00
	Mean	231.87	259.53	241.63
	F-value	37.05**	30.48**	62.82**
	LSD	47.45	55.64	51.03
	CV (%)	6.00	7.00	7.00
Spike length (cm)	Min	4.40	4.00	4.25
	Max	13.60	16.00	14.30
	Mean	9.17	10.96	10.07
	F-value	8.50**	7.36**	25.60**
	LSD	2.00	0.77	1.49
	CV (%)	7.00	2.00	5.00
Number of grains per spike	Min	10.3	11.00	10.65
	Max	48.70	63.00	55.85
	Mean	27.97	31.70	29.83
	F-value	33.02**	8.82**	16.91**
	LSD	4.11	10.90	8.13
	CV (%)	5.00	11.00	9.00
Grain weight per spike (g)	Min	0.40	0.50	0.45
	Max	3.3	3.0	3.15
	Mean	1.33	1.44	1.39
	F-value	16.83**	10.62**	21.57**
	LSD	0.43	0.31	0.47
	CV (%)	9.00	11.00	10.00
Grain yield (kg ha− 1)	Min	1028.00	1076.38	1052.94
	Max	4479.16	3840.27	4159.72
	Mean	1930.76	2126.05	2028.41
	F-value	8.75**	3.49*	6.12**
	LSD	820.64	1020.81	900.11
	CV (%)	13.00	15.00	14.00

CV Coefficient of variation. **P < 0.01

Fig. 3.

Phenotypic distribution of yield and yield-related traits in wheat genotypes evaluated across two growing seasons. The panels (a–g) show the variation in plant height, days to heading, spikes per m², spike length, grains per spike, grain weight per spike, and grain yield, respectively

Plant height differed significantly among genotypes in both years and in the combined analysis (P < 0.01; Table 2). The genotype × year interaction was also highly significant. Mean plant height increased markedly from 74.33 cm in 2013 to 95.19 cm in 2014, with an overall average of 84.76 cm (Fig. 3a). The tallest genotype across years was G106 (108.75 cm), while G118 was the shortest (56.25 cm). Ninety-seven genotypes exceeded the general mean, and 93 landraces had greater plant height than the average of the registered cultivars (89.17 cm).

The number of spikes per m² showed highly significant variation among genotypes, significant year effects, and a highly significant genotype × year interaction (P < 0.01; Table 2). The mean number of spikes per m² was 231.38 in 2013 and 259.00 in 2014 (overall mean 245.19; Fig. 3c). The highest value across years was recorded for G51 (489.67 spikes m⁻²), whereas the lowest was observed in G115 (87.67 spikes m⁻²). Among cultivars, Alparslan showed the highest spike density (433.00), while Ayyıldız had the lowest (216.00). Forty-two landraces exceeded the mean value of the registered cultivars (298.50).

Spike length was significantly affected by genotype, year, and genotype × year interaction (P < 0.01; Table 2). Mean spike length increased from 9.26 cm in 2013 to 11.02 cm in 2014 (Fig. 3d). The longest spike was recorded for G23 in the first year (13.70 cm) and G46 in the second year (16.15 cm). The shortest spikes were observed mainly in club wheat genotypes, particularly G61 (4.51 cm). The average spike length of the registered cultivars (10.73 cm) was slightly higher than that of the landraces (10.07 cm).

The number of grains per spike differed significantly among genotypes, years, and G×Y interaction (P < 0.01; Table 2). The overall mean increased from 27.93 grains in 2013 to 31.53 grains in 2014 (Fig. 3e). Ayyıldız had the highest combined value (51.35), while G32 had the lowest (16.51). Among landraces, only G81 (45.38) and G15 (44.32) exceeded the best registered cultivars. Grain weight per spike also showed highly significant genotypic differences and G×Y interactions (P < 0.01; Table 2), though the year effect was not significant. The mean grain weight per spike was 1.33 g in 2013 and 1.45 g in 2014 (Fig. 3f). The highest values were recorded for Ayyıldız (3.15 g), Palandöken 97 (3.00 g), and Alparslan (2.55 g). The best performing landrace was G121 (2.18 g). The overall mean of the landraces (1.35 g) was clearly lower than that of the registered cultivars (2.58 g).

Grain yield showed highly significant differences among genotypes and G×Y interactions (P < 0.01; Table 2), while the overall year effect was not statistically significant. In the first year, the highest yields were obtained from Ayyıldız (4319.40 kg ha⁻¹) and Doğu 88 (4046.30 kg ha⁻¹), followed by landraces G165, G180, G121, and G122. In the second year, Alparslan was the highest yielding cultivar (2361.10 kg ha⁻¹), although 59 landraces outperformed it. The gap between genotypes narrowed substantially in the second year; the mean yield of the landraces increased from 1472.20 to 1640.30 kg ha⁻¹, while the cultivars declined from 3508.50 to 1995.40 kg ha⁻¹. Across years, the best performing landrace was G165 (3524.90 kg ha⁻¹).

Quality traits

Analysis of variance showed that genotype, year, and genotype × year interaction effects for thousand kernel weight were all highly significant (P < 0.01; Table 3). The overall mean thousand kernel weight was 41.00 g, with values decreasing from 42.47 g in 2013 to 39.52 g in 2014 (Table 3). The highest thousand kernel weights were consistently recorded in the same group of landraces across both years. In 2013, genotypes 24, 22, 26, and 126 ranked first, with values of 58.17, 56.97, 54.67, and 52.95 g, respectively. In 2014, the same four genotypes remained the top performers, although their ranking slightly changed. Across years, genotype 24 showed the highest mean value (55.13 g), whereas Alparslan had the lowest mean thousand kernel weight among all entries (28.05 g). Landrace genotypes had higher mean thousand kernel weights than the registered cultivars in both years and in the combined analysis. The average values of the landraces were 42.69 g in 2013, 39.62 g in 2014, and 41.16 g across years, whereas the corresponding values for the registered cultivars were 35.92, 36.58, and 36.25 g, respectively. Among the cultivars, Ayyıldız had the highest thousand kernel weight (46.35 g). Overall, 160 landraces exceeded the mean of the registered cultivars (Table 3; Fig. 4a).

Table 3.

Descriptive statistics and Analysis of Variance (ANOVA) for yield and yield-related traits of the evaluated genotypes across two consecutive growing seasons

Trait	Statistic	Year 1	Year 2	Combined (Year 1 and Year 2)
1000-grain weight (g)	Min	27.50	26.90	27.20
	Max	53.40	50.60	52.00
	Mean	41.67	38.91	40.29
	F-value	17.92**	23.72**	34.72**
	LSD	3.99	3.18	3.56
	CV (%)	3.00	3.00	3.00
Hectoliter weight (kg L–1)	Min	78.80	75.00	74.90
	Max	85.40	85.40	85.40
	Mean	82.11	81.88	81.99
	F-value	4.32**	6.90**	9.21**
	LSD	4.12	3.22	3.65
	CV (%)	2.00	1.00	1.00
Zeleny sedimentation (mL)	Min	10.00	18.00	14.00
	Max	52.00	69.00	60.50
	Mean	33.55	36.08	34.82
	F-value	11.02**	16.00**	16.36**
	LSD	6.66	6.83	6.66
	CV (%)	6.00	6.00	6.00
Protein ratio	Min	12.00	11.70	11.85
	Max	16.1	15.50	15.95
	Mean	14.34	14.23	14.28
	F-value	3.72*	1.46	3.33**
	LSD	1.40	1.90	1.64
	CV (%)	3.00	4.00	4.00

CV Coefficient of variation. **P < 0.01

Fig. 4.

Phenotypic distribution of quality traits in wheat genotypes evaluated across two growing seasons. The panels (a–d) show the variation in 1000-grain weight (g), hectoliter weight (kg L^–1), Zeleny sedimentation (mL), protein ratio, respectively

Hectoliter weight differed significantly among genotypes in both years and in the combined analysis (P < 0.01; Table 3). In contrast, year and genotype × year interaction effects were not significant. The overall mean hectoliter weight was 81.58 kg hL⁻¹, with very similar yearly means of 81.68 kg hL⁻¹ in 2013 and 81.47 kg hL⁻¹ in 2014 (Table 3). The highest hectoliter weight in 2013 was recorded in Alparslan (85.40 kg hL⁻¹), whereas genotype 141 had the highest value in 2014 (85.43 kg hL⁻¹). The lowest hectoliter weight in both years was observed in genotype 6, with values of 71.30 and 72.35 kg hL⁻¹, respectively. In the combined analysis, 117 genotypes exceeded the overall mean. Among the registered cultivars, hectoliter weight ranked as follows: Gerek 79 (82.65 kg hL⁻¹), Palandöken 97 (81.60 kg hL⁻¹), Alparslan (81.10 kg hL⁻¹), Ayyıldız (80.40 kg hL⁻¹), Karahan 99 (79.00 kg hL⁻¹), and Doğu 88 (78.40 kg hL⁻¹). Only seven landraces had hectolitre weights below the suggested threshold value of 75 kg hL⁻¹ (Table 3; Fig. 4b).

Genotype, year, and genotype × year interaction effects for Zeleny sedimentation were all highly significant (P < 0.01; Table 3). Mean sedimentation increased from 33.30 mL in 2013 to 35.27 mL in 2014, with a combined mean of 34.28 mL (Table 3). In 2013, the highest sedimentation value was obtained from genotype 159 (51.67 mL), while the lowest was recorded in genotype 98 (10.00 mL). In 2014, genotype 60 had the highest value (68.50 mL), whereas genotype 44 had the lowest (17.50 mL). Across years, genotype 60 ranked first with a mean Zeleny sedimentation value of 56.92 mL, while genotype 165 had the lowest combined value (21.58 mL). The average Zeleny sedimentation of the registered cultivars was slightly lower than that of the landraces in 2013, but higher in 2014. Across years, however, the registered cultivars showed a slightly higher mean (35.33 mL) than the landraces (34.25 mL). Among the cultivars, Alparslan had the highest sedimentation value (51.00 mL), followed by Doğu 88 (38.00 mL), Ayyıldız (32.50 mL), Gerek 79 (32.00 mL), Palandöken 97 (30.50 mL), and Karahan 99 (28.00 mL). Based on the Turkish Food Codex requirement for bread wheat flour, 143 genotypes had sedimentation values above the threshold of 30 mL (Table 3; Fig. 4c).

Protein content showed significant genotypic differences in the first year (P < 0.05), non-significant differences in the second year, and highly significant differences in the combined analysis (P < 0.01; Table 3). Year and genotype × year interaction effects were not significant. The overall mean protein content was 14.27%, with nearly identical yearly means of 14.28% in 2013 and 14.27% in 2014 (Table 3). The highest protein content in 2013 was observed in genotype 104 (16.10%), while genotype 67 had the highest value in 2014 (15.98%). In the combined analysis, genotype 111 ranked first with 15.94%, whereas genotype 130 had the lowest value with 11.93%. A total of 95 genotypes exceeded the general mean of the experiment. The registered cultivars showed relatively high protein values in both years. Doğu 88 had the highest average protein content among the cultivars (15.35%), followed by Alparslan (15.10%), Ayyıldız (14.95%), Gerek 79 (13.65%), Palandöken 97 (13.55%), and Karahan 99 (12.75%). All genotypes exceeded the minimum protein content threshold of 10.5% defined by the Turkish Food Codex for bread wheat (Table 3; Fig. 4d).

Cold tolerance tests

Cold tolerance tests demonstrated highly significant differences among genotypes at − 13, −15, and − 17 °C (P < 0.01), and significant differences at − 19 °C (P < 0.05) (Table 4). The controlled freezing assay clearly separated the material into tolerant and susceptible groups. The contrasting phenotypic responses of genotypes to freezing stress, particularly in terms of survival and tissue damage, are clearly documented in Fig. 5, corroborating the results obtained from the controlled freezing assays. At − 13 °C, all registered cultivars and 28 landraces survived at acceptable levels. At − 15 °C, all registered cultivars except Gerek 79, together with seven landraces, remained viable. At − 17 °C, only Alparslan and Doğu 88 met the survival criterion, with Alparslan maintaining 100% survival and Doğu 88 retaining 60% live plants. At − 19 °C, only Alparslan remained clearly tolerant, with 80% survival. Although some genotypes such as 173, 81, and Ayyıldız still showed partial survival at − 17 °C, they did not meet the threshold of five living plants out of ten required for classification as tolerant at that temperature. Based on this criterion, Alparslan was classified as tolerant to − 19 °C; Doğu 88 to − 17 °C; Ayyıldız, Palandöken 97, Karahan 99, and landraces 173, 126, 15, 125, 171, and 132 to − 15 °C; and 15 additional landraces to − 13 °C. Most of the remaining landraces failed to survive − 13 °C adequately (Table 4; Fig. 6). These findings indicate that the registered cultivars bred for winter habit and regional adaptation possessed a clear advantage in freezing tolerance. Cultivars developed for Eastern Anatolia, such as Alparslan, Doğu 88, Ayyıldız, and Palandöken 97, outperformed cultivars developed for other regions, confirming the effectiveness of breeding for winter hardiness under severe continental conditions. By contrast, only 15% of the landraces survived − 13 °C and only 3.89% survived − 15 °C, suggesting that a large proportion of the landrace material may be facultative or spring type rather than strict winter wheat. Nevertheless, several landraces, especially genotypes 81, 125, 173, 126, 15, 171, and 132, showed promising levels of freezing tolerance and may serve as useful parental resources for breeding. Since winter hardiness is a key determinant of stand establishment, survival, and eventually yield in winter wheat, the identification of such tolerant landraces is particularly valuable for breeding programs targeting high altitude and cold prone environments.

Table 4.

Descriptive statistics and Analysis of Variance (ANOVA) for cold resistance

Temperature (°C)	Statistic	Surviving genotypes
-13	Min.	0.00
	Max.	10.00
	Total1	140.00
	F-value	148.21**2
	LSD	0.73
-15	Min.	0.00
	Max.	10.00
	Total	79.00
	F-value	10.02**
	LSD	2.37
-17	Min.	0.00
	Max.	10.00
	Total	26.00
	F-value	7.95**
	LSD	1.73
-19	Min.	0.00
	Max.	9.00
	Total	9.00
	F-value	6.14*
	LSD	1.29

¹ Total value represents the number of surviving genotypes following freezing treatment at each temperature threshold. ² * p < 0.05, ** p < 0.01

Fig. 5.

Images from cold resistance studies

Fig. 6.

Cold tolerance of the genotypes used in the experiment (⁰C)

Multivariate analysis

Cluster analysis

Agglomerative hierarchical clustering (AHC), based on Euclidean distances and Ward’s agglomeration method, was utilized to assess the phenotypic divergence among the 186 wheat genotypes. The resulting dendrogram optimally partitioned the germplasm into two major, distinct clusters (k = 2), reflecting a fundamental phenotypic dichotomy within the studied population. Although the germplasm was structured into partially overlapping groups, the centroid profiles demonstrated biologically meaningful differentiation between the two major clades. Cluster I constituted the majority of the population, comprising 107 genotypes (G1, G2, G4, G5, G6, G7, G8, G9, G10, G11, G13, G15, G16, G22, G23, G24, G25, G26, G27, G28, G29, G30, G33, G34, G35, G36, G37, G38, G39, G40, G41, G42, G43, G45, G46, G47, G49, G53, G54, G55, G56, G57, G58, G60, G62, G63, G64, G65, G66, G67, G68, G69, G70, G71, G72, G73, G74, G75, G76, G77, G78, G80, G81, G83, G84, G89, G90, G91, G93, G100, G104, G109, G110, G111, G112, G114, G115, G117, G118, G119, G120, G124, G125, G126, G127, G128, G129, G131, G133, G134, G135, G138, G142, G145, G147, G152, G156, G160, G166, G170, G171, G173, G174, G175, G176, G178, and G186), accounting for 57.53% of the total germplasm. This dominant group represented a broad phenotypic base within the studied population. Notably, the standard check variety Karahan 99 (G186) was grouped within this cluster, differentiating it from the other commercial checks. Cluster II contained a highly divergent subset of 79 genotypes (G3, G12, G14, G17, G18, G19, G20, G21, G31, G32, G44, G48, G50, G51, G52, G59, G61, G79, G82, G85, G86, G87, G88, G92, G94, G95, G96, G97, G98, G99, G101, G102, G103, G105, G106, G107, G108, G113, G116, G121, G122, G123, G130, G132, G136, G137, G139, G140, G141, G143, G144, G146, G148, G149, G150, G151, G153, G154, G155, G157, G158, G159, G161, G162, G163, G164, G165, G167, G168, G169, G172, G177, G179, G180, G181, G182, G183, G184, and G185), representing 42.47% of the population. The centroid profiles indicated that this cluster formed a phenotypically distinct phenotypic group compared to the main germplasm base in Cluster I. Interestingly, the other five established check varieties Doğu 88 (G181), Palandöken 97 (G182), Alparslan (G183), Ayyildiz (G184), and Gerek 79 (G185) clustered together in this group. This tight grouping indicates shared phenotypic and genetic affinities among these specific cultivars, distinctively separating them from the broader germplasm pool found in Cluster I (Fig. 7). The substantial Euclidean distance between the two primary cluster centroids indicates the presence of two major phenotypically differentiated groups within the evaluated germplasm.

Fig. 7.

Dendrogram graph of genotypes for cold resistance

Heatmap analysis

A clustered heatmap was generated to visualize the multivariate relationships among the wheat genotypes and measured traits. The analysis was performed using standardized trait values, and hierarchical clustering was applied to both genotypes and variables based on Euclidean distance and Ward’s linkage method. The heatmap revealed two major genotype groups (Cluster I and Cluster II), consistent with the hierarchical clustering results. Cluster I comprised genotypes characterized by relatively higher values for yield-related and quality-associated traits, whereas Cluster II included genotypes exhibiting comparatively lower or intermediate performance across most variables. The color gradient represented standardized trait values, where warmer colors indicated above-average values and cooler colors indicated below-average values. The dendrogram structure demonstrated clear trait co-associations, particularly among yield components and grain quality parameters, suggesting coordinated variation among these traits. The separation of genotypes into two major clusters indicates the presence of moderate phenotypic diversity within the germplasm panel (Fig. 8). Overall, the heatmap confirmed the clustering pattern observed in hierarchical clustering and provided additional insights into trait relationships, highlighting distinct genotype groups with contrasting agronomic and quality profiles. These findings support the use of selected genotypes from different clusters as potential parents for breeding programs aimed at maximizing genetic variability.

Fig. 8.

Clustered heatmap showing relationships among wheat genotypes and measured traits based on Euclidean distance and Ward’s method

Principal Component Analysis (PCA)

To further elucidate the multivariate relationships, trait associations, and underlying germplasm structure, a principal component analysis (PCA) was performed. The analysis revealed that the first two principal components (PC1 and PC2) accounted for 24.4% and 19.1% of the total phenotypic variation, respectively, cumulatively explaining 43.50% of the data’s total variance (Fig. 9). The correlation heatmap (Fig. 9) revealed a critical synergy: grain weight (GW) and grains per spike (GPS) maintained positive associations with early cold-survival classes (C1-C2), suggesting that reproductive efficiency can be improved without compromising winter hardiness. An evaluation of the variable loadings indicated that PC1 was primarily driven by strong positive contributions from cold tolerance parameters (C1, C2), grain weight per spike (GW), and grains per spike (GPS), representing a primary axis of winter hardiness and spike productivity. Conversely, PC2 was predominantly associated with grain yield (Y), spikes per m² (SPM), and plant height (PH), functioning as an axis of stand-level agronomic performance. The PCA biplot visually reinforced the phenotypic differentiation among the evaluated genotypes observed in the cluster analysis. While most of the genotypes formed a central, overlapping continuum indicating moderate phenotypic variation, a highly distinct subset of accessions occupied peripheral positions in the ordination space. Specifically, genotypes such as G181, G183, G184, G173, G81, G126, and G125 exhibited extremely positive scores along the first axis, confirming their superior cold tolerance and spike architectural traits. Among these, G183 emerged as the most divergent genotype with exceptionally high scores, followed by G181 and G184, clearly separating them from the central genotypic cloud. From a practical breeding perspective, these multivariate results underscore the presence of complementary genetic resources within the wheat germplasm evaluated. While the broader population provides an excellent foundation for improving yield, spike density, and hectoliter weight, the distinct sub-clade previously identified (encompassing G125, G126, G15, G173, G181, G183, G184, and G81) harbors exceptionally valuable alleles for freezing tolerance, superior individual spike performance (GW and GPS), and essential quality traits. Breeding for winter wheat improvement in cold-prone and climatically challenging regions, such as Eastern Anatolia, necessitates simultaneous progress in winter survival, yield stability, and end-use quality. The present analyses indicate that these multifaceted goals can be effectively achieved by strategic hybridization between contrasting multivariate groups. Crosses between the high-stand-productivity genotypes (central/PC2-driven group) and the elite cold-tolerant accessions (peripheral/PC1-driven group) are expected to generate highly favorable recombination’s. The distinct genotypes repeatedly identified by both clustering and PCA, most notably G183 due to its extreme multivariate uniqueness deserve special attention as highly promising parental materials for future introgression efforts, field validation, and targeted genomic characterization.

Fig. 9.

Correlation heat map of measured agronomic traits and cold tolerance variables providing the basis for the PCA and factor rotation

Correlation analysis

Pearson correlation analysis revealed significant and biologically meaningful associations among the evaluated agronomic, quality, and cold tolerance traits (Fig. 10). A highly strong positive correlation was observed between grain weight per spike (GW) and grains per spike (GPS) (r = 0.834). Both yield components also exhibited moderate positive associations with cold tolerance parameters, particularly C1 (r = 0.51 and 0.42, respectively) and C2 (r = 0.45 and 0.41, respectively). Grain yield (Y) was positively associated with by plant height (PH; r = 0.379), spikes per m² (SPM; r = 0.460), and hectoliter weight (HW; r = 0.362). Conversely, HW demonstrated significant negative correlations with GW (r = -0.543), GPS (r = -0.407), and thousand-grain weight (TGW; r = -0.365), highlighting a classic physiological trade-off between grain density and individual grain dimensions. Regarding quality parameters, protein content (P) and sedimentation value (S) were positively correlated (r = 0.372), reaffirming the established link between protein accumulation and bread-making quality.

Fig. 10.

Pearson correlation matrix and heat map displaying the relationships among the measured agronomic and physiological traits

Estimation of genetic merit using Best Linear Unbiased Prediction (BLUP)

The estimation of genetic merit based on best linear unbiased prediction (BLUP) revealed substantial variation among the evaluated wheat genotypes across all studied traits, indicating the presence of exploitable genetic diversity for breeding programs. The use of BLUP allowed the partitioning of environmental noise and provided more reliable estimates of genotypic performance compared to raw phenotypic means. The ranking of genotypes based on BLUP-derived values demonstrated clear superiority of specific genotypes for yield and associated agronomic traits. For instance, grain yield (Y) exhibited a wide range among genotypes, with several entries such as genotype 165, 180, and 184 showing markedly higher predicted values (> 3000 kg ha⁻¹), indicating strong genetic potential for productivity under the tested conditions. In contrast, lower-performing genotypes displayed substantially reduced BLUP estimates, reflecting either poor genetic potential or sensitivity to environmental variation. Similarly, important yield components such as spike length (SL), grain weight per spike (GW), and grains per spike (GPS) also showed considerable variability. High-performing genotypes generally combined favorable values across these traits, suggesting a cumulative effect contributing to overall yield superiority. For example, genotypes with high GW and GPS values (G 182, G183, and G184) also tended to exhibit superior yield performance, highlighting the strong association between these components and final productivity. Quality-related traits, including protein content (P), thousand grain weight (TGW), and hectoliter weight (HW), also exhibited distinct genotypic differences. Some genotypes combined high yield with acceptable or superior quality traits, which are particularly important for simultaneous improvement of productivity and end-use quality. For instance, genotypes such as 181 and 183 showed relatively high protein content alongside competitive yield performance, suggesting their suitability for breeding programs targeting both yield and quality improvement. Cold tolerance indices (C1–C4) further differentiated genotypes in terms of stress adaptation. While many genotypes showed near-zero or negative BLUP values for these indices, a subset of genotypes (G181, G183, and G184) exhibited notably higher values, indicating enhanced tolerance to low-temperature stress conditions. This variation suggests the possibility of selecting genotypes combining both high yield potential and abiotic stress resilience. Overall, the BLUP-based evaluation enabled the identification of superior genotypes with consistent performance across multiple traits. The results emphasize the effectiveness of BLUP in capturing true genetic merit and support its application in multi-trait selection strategies. The observed variability among genotypes provides a strong foundation for future breeding efforts aimed at improving yield, quality, and stress tolerance simultaneously.

Multi-trait selection and identification of ideal genotypes using the MGIDI index

The KMO value indicated acceptable sampling adequacy for factor analysis, while Bartlett’s test confirmed significant correlations among variables suitable for dimensional reduction (P < 0.001). All agro-morphological and yield-related traits were utilized to identify both cold-tolerant and cold-susceptible winter wheat genotypes. Genotype ranking was performed using the Multi-Trait Genotype–Ideotype Distance Index (MGIDI), which quantifies the distance between each genotype and an a priori defined ideotype (Table S3). Principal Component Analysis (PCA) retained five components (eigenvalues > 1) under cold stress, explaining 54.24% of the total phenotypic variability. The individual contributions of the retained factors were FA1 = 14.35%, FA2 = 13.61%, FA3 = 12.35%, FA4 = 7.07%, and FA5 = 6.85% (Table 5). The scree plot and cumulative variance supported the retention of these five components for the computation of the multi-trait index (Fig. 11). The factor-wise contribution analysis demonstrated distinct patterns, successfully grouping highly correlated traits. Specifically, FA1 was mainly associated with cold tolerance indices (C1, C2, C3, and C4), while FA2 represented yield and plant architecture components (grain yield, plant height, and spikes per square meter). FA3 included spike yield traits (grain weight and grains per spike), whereas FA4 captured grain physical properties and spike size (thousand-grain weight and spike length), and FA5 was loaded by specific phenological traits (protein content; P and sedimentation; S). The MGIDI ranking plot revealed a clear differentiation among the 186 winter wheat genotypes, with the selected genotypes positioned closer to the ideotype, indicating superior overall phenotypic performance and cold resilience. Based on the selection intensity, 28 genotypes were identified as superior candidates for multi-trait improvement. These selected genotypes included G183, G181, G81, G184, G173, G182, G122, G111, G152, G126, G171, G15, G11, G186, G79, G12, G125, G132, G116, G166, G59, G162, G14, G146, G48, G103, G185, and G114. The view of strengths and weaknesses provided critical insights into the specific trait contributions for each selected genotype. Genotypes with lower contribution percentages from specific factors indicate that those trait groups are already close to the ideal state (strengths). Conversely, higher contributions in some selected genotypes indicate residual gaps separating them from the ideotype, representing opportunities for further targeted improvement. The MGIDI approach effectively balanced these five trait groups, proving highly effective in identifying genotypes that combine high yield potential with desirable agronomic characteristics. Overall, the selected 28 genotypes, being closest to the ideotype, represent highly promising parental candidates for breeding programs aimed at the simultaneous improvement of yield, agronomic stability, and cold resilience in winter wheat.

Table 5.

BLUP-adjusted values of key agro-morphological traits and cold tolerance indices for the superior winter wheat genotypes selected by the MGIDI index

Genotype	Grain yield	Plant height	Thousand grain weight	C1	C2
G11	1937.66	100.00	48.87	10.00	0.99
G12	2173.60	105.00	39.42	8.00	-0.01
G14	2117.82	87.50	40.92	7.00	-0.01
G15	1850.46	80.00	44.52	10.00	4.99
G48	2121.41	87.50	40.42	0.00	-0.01
G59	2479.05	107.50	42.22	1.00	-0.01
G79	2052.18	85.00	41.57	5.00	0.99
G81	1444.49	95.00	39.27	10.00	9.99
G103	2638.88	97.50	40.27	0.00	-0.01
G111	1607.46	90.00	37.87	1.00	-0.01
G114	1989.68	80.00	41.52	0.00	-0.01
G116	2263.88	105.00	42.32	0.00	-0.01
G122	3093.85	102.50	45.72	1.00	0.99
G125	1545.13	72.50	47.52	8.00	7.99
G126	1857.68	82.50	51.92	10.00	5.99
G132	2107.74	80.00	42.82	6.00	4.99
G146	2111.10	85.00	38.22	0.00	-0.01
G152	1760.41	102.50	37.97	0.00	-0.01
G162	2239.63	85.00	40.92	0.00	-0.01
G166	1447.68	92.50	41.22	1.00	-0.01
G171	1763.66	87.50	40.07	6.00	4.99
G173	1833.49	80.00	42.07	10.00	6.99
G181	3063.99	87.50	28.72	10.00	8.99
G182	2867.38	100.00	41.47	9.00	3.99
G183	2832.57	82.50	27.37	10.00	9.99
G184	3423.57	85.00	46.97	10.00	6.99

Fig. 11.

Polar plot illustrating the ranking of genotypes in ascending order based on the Multi-trait Genotype–Ideotype Distance Index (MGIDI). Each point represents an evaluated genotype, with selected genotypes highlighted in red in the electronic version of the article. The red circle represents the cut point established according to the applied selection pressure, separating genotypes retained for selection from those not selected. This visualization facilitates the identification of superior genotypes exhibiting desirable multi-trait performance

Comparative multi-trait selection via FAI-BLUP index

To cross-validate the efficiency and robustness of the ideotype-oriented selection, the Factor Analysis and Ideotype-Design based on Best Linear Unbiased Prediction (FAI-BLUP) index was computed parallel to the MGIDI framework. The multi-trait genotype ranking under the primary breeding objective (Ideotype 1; ID1) which targeted maximized grain yield and quality components alongside minimized freezing injury scores demonstrated an exceptional convergence with the MGIDI outcomes. Based on the spatial probability of genotypes approaching the multi-trait ideotype, the FAI-BLUP index effectively discriminated against elite germplasm from the broader population continuum. The registered cultivars G184 (Ayyıldız), G183 (Alparslan), and G181 (Doğu 88) occupied the top three ranks within the FAI-BLUP framework, mathematically reinforcing their evolutionary advantage, phenotypic stability, and superior breeding value under extreme continental freezing thresholds. Crucially, specific landrace genotypes, most notably the highly divergent line G81, followed by G122, G152, G111, and G11, were simultaneously retained within the top tier of the selection pool (Table 6). This strong alignment between two independent multivariate selection models confirms that the integration of BLUP-derived genetic merit with factor-based distance indices minimizes the risk of selection bias under high genotype-by-environment interactions. Consequently, these convergent genotypes represent the most reliable parental candidates for transferring synchronous cold resilience and sink capacity into modern wheat breeding pipelines.

Table 6.

Ranking of the top 28 superior winter wheat genotypes selected by the FAI-BLUP index (Ideotype 1) based on multi-trait spatial probability

Rank	Genotype	Rank	Genotype	Rank	Genotype	Rank	Genotype
1	G184	8	G11	15	G121	22	G15
2	G183	9	G79	16	G186	23	G12
3	G181	10	G182	17	G180	24	G113
4	G81	11	G146	18	G114	25	G166
5	G122	12	G173	19	G36	26	G14
6	G152	13	G185	20	G162	27	G13
7	G111	14	G59	21	G116	28	G174

Strengths and weaknesses of the selected genotypes

The strengths and weaknesses of the 28 selected winter wheat genotypes were visualized based on the proportion of each factor’s contribution to the computed MGIDI index under cold stress conditions (Fig. 12). This radar chart provides critical insights into the specific trait groups that are already aligned with the ideal condition (i.e., strengths, represented by smaller proportional contributions plotted toward the external edge), and the trait groups that still separate the genotypes from the ideotype (weaknesses, represented by larger contributions drawn inward). The dashed line marks the theoretical value if all factors had contributed equally. The contribution of each factor varied notably among the selected genotypes, revealing distinct adaptive strategies to cold stress. For many of the superior genotypes, FA1 (cold tolerance indices) showed smaller contributions, indicating that their intrinsic cold resilience is already a major strength and closely matches the ideotype. Similarly, distinct genotypes exhibited strong performance (low contributions) in FA2 (yield and plant architecture) and FA3 (spike yield components), demonstrating their ability to maintain high grain yield and spike productivity despite freezing temperatures. Conversely, factors with larger contributions represent residual gaps that require further attention. For instance, certain selected genotypes displayed higher contributions in FA4 (grain physical properties and spike size) and FA5 (grain quality traits). This indicates that while these specific genotypes possess excellent cold tolerance and overall yield potential, there are still opportunities to further improve their thousand-grain weight and biochemical quality (protein and sedimentation) through targeted breeding. This indicates that while these specific genotypes possess excellent cold tolerance and overall yield potential, there are still opportunities to further improve their thousand-grain weight and phenological adaptability through targeted breeding. Ultimately, the strengths and weaknesses confirm that the MGIDI approach successfully identified wheat landraces and modern cultivars that balance cold resilience with superior agronomic performance.

Fig. 12.

Strengths and weaknesses of the superior winter wheat genotypes selected by the Multi-trait Genotype–Ideotype Distance Index (MGIDI), based on the proportional contribution of each latent factor (FA1–FA5). Lower factor contributions indicate traits closer to the ideotype and therefore represent the major strengths of the selected genotypes, whereas higher contributions indicate residual gaps from the ideotype. The dashed line represents the expected contribution if all factors contributed equally to the MGIDI. Factor composition: FA1, cold tolerance indices; FA2, yield and plant architecture; FA3, spike yield components; FA4, grain physical properties and spike size; and FA5, grain quality traits

Ranking analysis of superior winter wheat genotypes based on the MGIDI index was performed across five latent factors, as summarized in Table 7. This ranking reflects the relative performance of each genotype within individual factors under the natural environmental conditions of Erzurum. The results revealed substantial variability among genotypes in their contributions to each factor, reflecting their adaptation to high-altitude winter conditions. Notably, Genotype G94 achieved the highest rank in both FA1 (Tolerance indices) and FA5 (Grain quality traits), demonstrating an exceptional and consistent performance across both adaptive resilience and quality dimensions. In FA2 (Yield and plant architecture), G27 emerged as the top performer, while G95 ranked highest in FA3 (Spike yield components). Furthermore, G102 ranked first in FA4, highlighting its superiority in grain physical properties and spike size. Consistency across multiple factors was observed in several genotypes; for instance, G95 ranked first in FA3 and secured a place in the top five for both FA1 and FA2, suggesting a highly balanced performance profile for both yield and environmental adaptation in Erzurum. By contrast, G102, despite leading in FA4, did not appear in the top ten for FA1 or FA2, indicating that its performance was more specifically focused on grain size traits rather than overall stability. Other genotypes, such as G93 and G92, consistently achieved high rankings across various factors, representing stable and promising candidates for integrated breeding objectives in similar agro-ecosystems.

Table 7.

Ranking of superior genotypes for each factor

Rank	FA1	FA2	FA3	FA4	FA5
1	94	27	95	102	94
2	92	93	93	104	144
3	83	91	96	106	7
4	167	181	97	31	15
5	95	95	53	13	60
6	93	17	91	26	18
7	154	74	26	95	75
8	106	89	29	107	151
9	113	134	164	29	125
10	109	94	143	108	120
11	30	6	108	112	147
12	107	79	139	157	14

Factor loadings, communalities, and unexplained variance of traits across factors (FA1–FA5)

Factor loadings and communality analyses were conducted based on the MGIDI to evaluate the relationship between various agro-morphological traits and factors FA1 to FA5, as well as the proportion of variance explained by these factors. The results, including factor loadings for each trait across the five factors, communality, and unexplained variance (uniqueness), are presented in Table 8. These results emphasize the contribution of each trait to the factor structure and the explanatory power of the factors in accounting for trait variance under Erzurum’s environmental conditions. FA1 was strongly associated with adaptation and tolerance indices; C3 (0.939), C2 (0.699), C4 (0.681), and C1 (0.433) exhibited high factor loadings, indicating a key role for FA1 in explaining variance in environmental response. In contrast, FA2 was mainly associated with productivity and architecture, where grain yield (Y: 0.754), plant height (PH: 0.604), and hectoliter weight (HW: 0.586) showed high factor loadings. FA3 was linked to spike productivity, dominated by grain weight (GW: 0.922) and grains per spike (GPS: 0.826). Furthermore, FA4 was primarily associated with grain size and physical structure, showing high loadings for thousand grain weight (TGW: 0.847) and spike length (SL: 0.331). This suggests that SL is governed by distinct genetic mechanisms, offering an opportunity for independent selection in breeding programs. Finally, FA5 was strongly associated with grain biochemical quality, specifically sedimentation (S: 0.672) and protein content (P: 0.494). The communalities indicate the proportion of variance for each trait explained by the five factors. GW (0.995), GPS (0.785), and TGW (0.781) had the highest communalities, demonstrating the strong explanatory power of these factors for these specific productivity traits. In contrast, SL (0.142) show the lowest communalities, indicating higher unexplained variance (uniqueness) for these traits (0.857). Based on the factor structure, traits with the highest absolute loadings were considered the primary contributors to each factor, effectively categorizing the 15 evaluated traits into five distinct dimensions of wheat performance.

Table 8.

Factor loadings, communalities, and uniqueness of the evaluated wheat traits based on the Multi-Trait Genotype-Ideotype Distance Index (MGIDI)

Traits	FA1	FA2	FA3	FA4	FA5	Communality	Uniqueness
SL	0.042	0.048	0.047	0.330	0.162	0.142	0.857
GW	0.277	-0.142	0.922	0.196	-0.092	0.995	0.005
GPS	0.247	-0.173	0.826	-0.106	0.009	0.785	0.214
PH	-0.057	0.603	-0.149	0.178	0.076	0.427	0.572
Y	0.142	0.754	0.011	-0.121	-0.274	0.679	0.320
SPM	0.050	0.532	-0.108	-0.086	-0.005	0.305	0.694
SD	-0.003	-0.545	0.130	-0.135	-0.283	0.413	0.586
P	0.192	-0.109	-0.064	0.101	0.494	0.307	0.692
TGW	-0.146	-0.148	0.032	0.847	-0.135	0.781	0.218
HW	-0.166	0.586	-0.362	-0.291	0.224	0.638	0.361
S	0.060	0.160	-0.024	-0.043	0.671	0.483	0.516
C1	0.433	-0.202	0.356	0.079	-0.175	0.392	0.607
C2	0.699	0.002	0.272	0.024	0.016	0.564	0.435
C3	0.939	0.169	0.180	-0.087	0.209	0.995	0.004
C4	0.681	0.110	0.059	-0.103	0.242	0.550	0.449

Multi-trait selection gains and genetic evaluation

The selection gains (SG) and the genetic parameters of the evaluated traits for the 28 superior wheat genotypes selected by the MGIDI index are summarized in Table 9. The results demonstrate that the MGIDI-based selection effectively prioritized genotypes with enhanced performance across multiple trait categories. Substantial genetic progress was observed for key productivity and adaptation traits. Grain yield (Y) exhibited a significant selection gain of 6.13%, while other yield components such as spikes per square meter (SPM) and spike density (SD) reached gains of 11.23% and 5.90%, respectively. These positive gains confirm the robustness of the index in identifying high-yielding candidates within a diverse population. Furthermore, the selection provided desirable advancements in environmental resilience and quality; Tolerance index 3 (C3) showed a gain of 5.81%, while grain quality attributes such as protein content (P) and sedimentation (S) achieved gains of 0.91% and 3.61%, respectively. The heritability (h²) values ranged from 0.14 for spike length (SL) to as high as 0.99 for Grain weight (GW) and grain per spike (GPS). High heritability for major yield components suggests that a significant portion of the observed variation is genetically determined, ensuring reliable transmission of these traits to subsequent generations. Notably, thousand grain weight (TGW) also showed a strong heritability of 0.78 with a selection gain of 1.45%. The integration of these diverse traits into the MGIDI framework allowed for a balanced genetic advance. By accounting for the strengths and weaknesses of each genotype across five latent factors, the selection process effectively identified 28 genotypes that combine superior yield potential with high-altitude adaptation and favorable grain quality. These genotypes represent highly promising parental material for winter wheat breeding programs aimed at optimizing multi-trait performance in similar agro-ecosystems.

Table 9.

Selection Gains (SG) and genetic parameters for 15 traits in winter wheat genotypes selected by the MGIDI index

Trait	Xo	Xs	SD	h2	SG	SG (%)
Y	2022.20	2125.61	103.41	1.00	123.94	6.13
SPM	256.47	308.88	52.41	0.55	28.79	11.23
SD	163.63	172.96	9.32	1.00	9.65	5.90
C3	-0.08	0.00	0.08	1.00	-0.01	5.81
S	35.28	36.63	1.35	0.94	1.27	3.61
P	14.30	14.43	0.13	1.00	0.13	0.91
SL	10.07	10.10	0.03	0.14	0.00	0.04
GW	1.41	1.60	0.20	0.99	0.20	13.91
GPS	30.35	31.26	0.91	0.99	0.90	2.96
PH	84.74	79.82	-4.92	0.88	-4.33	-5.11
TGW	41.17	44.68	3.50	0.17	0.60	1.45
HW	81.59	78.75	-2.84	1.15	-3.27	-4.01
C1	2.04	4.75	2.71	1.00	2.71	132.49
C2	0.56	0.34	-0.21	1.00	-0.21	-38.21
C4	-0.11	-0.16	-0.05	1.00	-0.05	43.37

Xo Population mean, Xs Mean of selected genotypes, SD Selection differential, h² Broad-sense heritability, SG Selection gain, SG (%) Selection gain in percentage

Discussion

The results of the present study provide compelling evidence that winter wheat adaptation in high-altitude continental environments is primarily governed by the interaction between cold tolerance mechanisms and yield-forming processes. The present study revealed highly significant genotypic differences (P < 0.01) for all major agronomic, quality, and cold tolerance traits, confirming the presence of substantial exploitable genetic variability. These findings are consistent with previous studies conducted under semi-arid and continental environments, where phenotypic diversity is often amplified by fluctuating climatic conditions [38]. A major outcome of this study was the pronounced G×E interaction observed for key traits, including heading time, plant height, spike density, and grain yield. The earlier heading observed in the second growing season, coupled with increased values for yield components such as grains per spike and spike length, suggests that improved moisture availability and favorable temperature regimes accelerated plant development and enhanced reproductive success. This observation reinforces the well-established concept that phenological plasticity is a critical adaptive mechanism, enabling genotypes to escape or mitigate environmental stress during highly vulnerable developmental stages such as stem elongation and grain filling [39]. The delayed heading observed during the first growing season was likely associated with reduced spring rainfall and stronger diurnal temperature fluctuations, particularly subzero night temperatures, which may have slowed reproductive development and prolonged vegetative growth. The highly significant genotype × year interactions observed across most agronomic traits indicate that while these traits are strongly genotype-dependent, seasonal conditions heavily modulate their expression. Similar year- and genotype-dependent differences in heading time have been widely reported in recent wheat studies, highlighting that phenological plasticity is primarily driven by climatic variation and genotype-specific environmental interactions [25, 40, 41]. This wide phenotypic variation confirms that heading time is a stable yet environmentally sensitive genotypic trait, reflecting diverse adaptive strategies within the germplasm driven by both natural and artificial selection processes. Conversely, recent multi-environment evaluations demonstrate that uniform environmental constraints such as anomalous thermal profiles or compressed spring transition periods can temporarily mask this genotypic variability, leading to synchronized phenological responses across otherwise diverse populations [42].

The marked increase in plant height observed during the second year is closely associated with higher seasonal precipitation, particularly during the critical May window. Enhanced moisture availability directly promoted stem elongation and extended vegetative growth a positive environmental stimulus consistent with recent multi-environment evaluations in semi-arid wheat systems [43, 44]. Although increased plant stature offers dual-purpose advantages for straw yield in mixed crop-livestock systems, excessive height often exacerbates lodging susceptibility under optimal moisture. In the present study, however, lodging was entirely absent, likely due to the overarching low-humidity conditions and negligible disease pressure characteristic of the Eastern Anatolian plateau. Spike density and architectural dimensions were similarly modulated by inter-annual climatic fluctuations. The restricted spike density recorded in the first year was primarily driven by early-season moisture deficits, which severely constrained tillering and subsequent spikelet differentiation. Nevertheless, the highly stable ranking of elite versus low-performing genotypes across contrasting seasons confirms a robust genetic governance over tillering capacity, even as its ultimate magnitude is environmentally calibrated [45]. The elongation of spikes and increased grain numbers in the second year reflect optimized moisture supply during the highly sensitive stem elongation and spikelet fertility phases. Notably, the pronounced year-to-year fluctuation in spike length among the landraces demonstrates a higher degree of phenotypic plasticity and environmental responsiveness compared to the buffered, stable profiles of modern registered cultivars [46]. Regarding sink capacity, the genotypic response for grain number per spike was highly variable; while certain commercial lines maintained stable grain sets, specific divergent landraces exhibited an increase of nearly 50% under ameliorated conditions. This profound differential plasticity underscores distinct developmental timings and stress-escape strategies embedded within the germplasm [39].

Grain weight per spike emerged as one of the primary phenotypic discriminators between traditional landraces and modern improved cultivars. While certain landraces exhibited marked increases in grain-filling capacity under the favorable seasonal conditions of the second year, the registered cultivars demonstrated superior phenotypic stability and a more efficient assimilate partitioning system. This buffering capacity suggests that modern varieties have been optimized for a more consistent grain-filling trajectory, regardless of inter-annual climatic volatility. The superior performance of these cultivars reflects long-term breeding efforts targeting sink strength the metabolic capacity of the grain to effectively attract and store photo-assimilates—and the structural synchronization of grain filling [47]. Recent advances in wheat physiology confirm that this optimized sink-source dynamic is a prerequisite for stabilizing yield in climate-stressed regions, where the duration of the grain-filling period is often truncated by rising temperatures [46]. Ultimately, seasonal differences in mean grain yield were masked by large genotype-specific responses. The general superiority of registered cultivars over most landraces indicates the cumulative effect of long-term breeding for adaptation and stand productivity under Eastern Anatolian conditions. While the first year favored winter cultivars through better pre-winter establishment, the second year characterized by delayed emergence and irregular rainfall distribution highlighted the unique adaptive flexibilities of select landraces. These findings confirm that while improved cultivars maintain a clear yield advantage, specific landraces possess adaptive plasticity that renders them highly competitive under stress-prone continental environments [48].

An important finding of the present study was the superior performance of several modern cultivars relative to local landraces in both yield and freezing tolerance. While landraces are universally viewed as the primary reservoirs of stress-tolerance genes [49], our results revealed a significant yield gap, with registered cultivars (Alparslan, Ayyıldız) outperforming landraces by nearly 77% in the first year. This superiority was primarily driven by higher spike density (SPM) and grains per spike (GPS), confirming that modern selection has maximized stand-level agronomic performance [50]. However, a physiological trade-off was evident: while cultivars dominated in grain number, landraces exhibited significantly higher thousand-grain weight (TGW) (mean 41.16 g vs. 36.25 g). This inverse relationship represents a classical sink–source physiological constraint [51]. The superior TGW in landraces suggests they possess elite alleles for sink capacity. This may reflect stronger assimilate partitioning efficiency and prolonged grain filling capacity in specific landrace backgrounds under stress-prone environments. Historically, introgression of such alleles from landrace populations is hindered by linkage drag, where favorable sink capacity traits are tightly coupled with undesirable architectural flaws like tall stature or lodging susceptibility. Our findings suggest that continuous, targeted breeding efforts can successfully uncouple these traits, allowing for the combination of high TGW and modern stand architecture without a genetic penalty.

The controlled freezing assays clearly differentiated genotypes based on their survival thresholds, revealing that winter hardiness in this germplasm follows a strict threshold model. The cultivar Alparslan’s ability to survive at − 19 °C establishes it as a robust benchmark for extreme cold resilience. This resilience may be associated with physiological mechanisms such as membrane stabilization, osmotic adjustment, and improved reactive oxygen species scavenging under freezing stress [52]. Interestingly, only 15% of landraces survived at − 13 °C, suggesting that many traditional genotypes in the region might be facultative or spring types rather than strict winter types, relying on stress-escape mechanisms rather than deep-freezing tolerance. This challenges the conventional assumption that landraces universally possess superior stress tolerance and highlights the success of modern, multi-decadal regional breeding programs in fixing adaptive alleles for cold resilience. Nevertheless, specific landraces (G81, G173) maintained viability at − 15 °C while retaining superior quality traits, representing divergent genetic resources that bridge the gap between landrace-driven grain quality and cultivar-driven cold resilience.

Future climate models predict increasingly mild winters at higher latitudes, which paradoxically complicates selection due to the sporadic incidence of winter injury, loss of snow cover insulation, and fluctuating environmental stresses like late-spring frosts [53]. In such snow-dominant regions, accurately assessing genetic merit beneath inconsistent cover remains a profound statistical challenge. This highlights the critical importance of BLUP in breeding programs, as it enables the identification of true genetic merit independent of environmental variability, particularly under unstable climatic conditions. Our study demonstrates that employing BLUP-based estimates is vastly superior to relying on raw phenotypic means. Unlike traditional selection approaches, MGIDI allows the simultaneous optimization of multiple correlated traits, thereby enabling the identification of balanced ideotypes rather than trait-specific extremes. By summarizing multi-trait variation into five latent factors, MGIDI effectively resolved issues of multicollinearity. The achieved selection gains specifically 6.13% for grain yield and 11.23% for spikes per square meter outperform traditional indices in simultaneous trait improvement. To further validate the robustness of this multi-trait selection, the FAI-BLUP index [54] was integrated into our framework. The exceptional convergence between MGIDI and FAI-BLUP in identifying the exact same elite ideotypes (G184, G183, G181, and the highly divergent landrace G81) mathematically confirms the stability of these genotypes across diverse spatial probability scenarios. Utilizing dual ideotype-design approaches effectively eliminates selection bias and solidifies the identification of true climate-resilient germplasm. Although the experimental data were originally collected during the 2013–2015 growing seasons, the increasing frequency of unstable snow cover and erratic winter temperatures observed over the last decade has further increased the relevance of baseline winter hardiness evaluations in continental wheat germplasm.

Our multivariate analyses strongly suggested that spike productivity traits (FA3: GW and GPS) were not necessarily negatively associated with cold hardiness (FA1), indicating the possibility of simultaneous improvement for both adaptation and productivity traits. Furthermore, a strategically important finding was the extraordinarily high uniqueness of spike length (SL; 0.857). In standard breeding populations, spike architecture is almost always pleiotropically linked to primary yield components. However, our data explicitly suggests that in this germplasm, SL remains a genetically standalone trait. This independence allows breeders to modify spike architecture—such as selecting for longer, looser spikes to improve canopy light interception and reduce micro-humidity—without unintended negative impacts on grain density or winter survival. Beyond agronomic performance, grain quality traits, including protein content (14.27%) and sedimentation value, showed highly stable expression despite environmental variability. Cold stress frequently disrupts nitrogen remobilization from vegetative tissues to the developing grain [55]. However, the moderate yet consistent selection gains observed for protein content (0.91%) and sedimentation value (3.61%) indicate that improvements in yield and stress tolerance did not compromise end-use quality. The ability of elite commercial lines to maintain stable profiles implies highly efficient nitrogen transport and partitioning mechanisms, resulting in premium grain suitable for demanding industrial baking processes [56].

Taken together, these findings highlight that combining BLUP-based genetic evaluation with dual multi-trait selection indices (MGIDI and FAI-BLUP) provides a highly efficient framework for identifying superior wheat genotypes under cold-prone environments. The factor contribution analysis further demonstrated that FA1 contributed predominantly to freezing tolerance and adaptation traits, whereas FA2 and FA3 mainly captured productivity-related variation. The identified elite genotypes represent valuable parental resources for future breeding programs aiming to simultaneously improve yield, winter survival, and grain quality under increasing climatic uncertainty. A key novelty of this study lies in the integration of large-scale landrace germplasm with advanced ideotype-based selection indices under extreme continental conditions, providing a scalable framework for climate-resilient wheat breeding. Hybridization between the genetically divergent cultivar-driven group and elite landrace-derived genotypes may provide promising opportunities for broadening the genetic base and improving cold adaptation [57, 58]. It should be noted that the present study was conducted under a single high-altitude continental environment across two growing seasons. Therefore, the observed adaptation patterns and genotype performance should be interpreted within the environmental scope of Eastern Anatolian conditions, and broader adaptation inferences require further multi-location validation.

Conclusion

This study provides a comprehensive evaluation of agronomic performance, grain quality, and freezing tolerance within a diverse winter wheat germplasm panel under the harsh continental conditions of Eastern Anatolia. By integrating BLUP-based estimates with the MGIDI multi-trait selection framework, the study enabled a robust identification of superior genotypes while minimizing environmental noise and simultaneously considering multiple correlated traits. Significant genotype × environment interactions confirmed that seasonal climatic variability strongly influenced phenology, yield formation, and stress adaptation. Modern registered cultivars, particularly Alparslan, Doğu 88, and Ayyıldız, generally exhibited superior grain yield, stand productivity, and winter hardiness, highlighting the effectiveness of long-term regional breeding for adaptation to severe winter environments. However, several landrace genotypes, including G15, G81, G121, G125, G126, G132, G171, and G173, retained valuable adaptive characteristics, especially for thousand-kernel weight, grain quality, and partial freezing tolerance under extreme low-temperature stress. These findings demonstrate that although modern cultivars maintain a clear agronomic advantage, elite landraces remain important reservoirs of adaptive genetic variation and physiological plasticity. Multivariate analyses further revealed that specific trait combinations associated with yield, cold tolerance, and grain quality can be simultaneously improved through ideotype-based selection strategies. Therefore, the strategic utilization of elite landraces together with high-performing cultivars offers a promising pathway for developing climate-resilient winter wheat ideotypes adapted to cold-prone environments. Future integration of high-throughput phenotyping with genomic approaches such as GWAS and QTL mapping will further accelerate the identification of key adaptive loci and support the development of next-generation winter wheat cultivars under increasing climatic uncertainty.

Authors contributions

Conceptualization, Ü.K. and M.T.; methodology, Ü.K., M.T., B.D., and A.T.; software, H.A. and A.T.; validation, B.D. and A.T.; formal analysis, H.A. and A.T.; investigation, Ü.K., B.D., and A.T.; resources, Ü.K. and M.T.; data curation, A.T.; writing—original draft preparation, H.A., A.T., and Ü.K.; writing—review and editing, A.T., M.T., B.D., and Ü.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the General Directorate of Agricultural Research and Policies (TAGEM), grant number TAGEM/TBAD/15/A12/P01/001, under the project titled “Determination of the Performance and Cold Resistance Levels of Wheat Landraces from Eastern Anatolia Under Erzurum Conditions”. This manuscript is partly derived from a PhD thesis completed in 2016.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.