Introduction

Durum wheat is the second most cultivated Triticum species, it is possible that it had greater importance in the past, but its extent is unknown (Martínez-Moreno et al. 2022). World durum production fluctuated significantly in the last decade between 31.4 and 38.7 million tonnes, however, it is alarming that in the last five years actual and forecasted world consumption surpasses world production (European Commission 2024). North America is the biggest producer of durum wheat, in particular Canada as global top producer with 5.4 million metric tonnes. In Europe, Italy is the main producer with 3.7 million metric tonnes. Other important producers are Turkey (2.9 million mt), Algeria (2.8 million mt) and India (1.1 million mt) (Euronext 2023). Durum wheat is well adapted to the Mediterranean area (Royo et al. 2014), but its productivity are limited by abiotic stresses such as water stress and high temperatures (Xynias et al. 2020). Marti and Slafer (2014) suggested that modern durum wheat varieties can exceed bread wheat in better growing areas. An increasing interest in durum wheat has been observed in recent years worldwide due to population growth, changes in dietary habits and in favour of the healthy Mediterranean diet where low-cost, convenient, versatile, and nutritious durum products with a long shelf-life such as dry pasta, couscous and bulgur play a major role (Webb 2019; Hammami et al. 2022; Shah et al. 2022). Durum wheat and the food made of it is not only an important source of energy but also rich in vitamins (B, E), minerals (potassium, magnesium, iron, folic acid) and an essential nutrient in human diet (Lintas 1988; Grant et al. 2012; Saini et al. 2023). The strong gluten structure is able to retain starch molecules during processing and cooking (Feillet 1984), thus the surface of the pasta does not become sticky or slimy and preserves its shape solidly (Dexter and Matsuo 1980). Durum wheat has meanwhilst found its place also in Hungarian crop production. In 2022, its growing area was around 35,000 ha, which was almost 4% of the wheat growing area. Along with the growing interest in durum wheat in the world and in Hungary, the proportion of organic areas in Europe is steadily increasing (Willer et al. 2023). According to several studies, organic farming could be one of the tools to a more sustainable farming system (Bux et al. 2022; Ingraffia et al. 2022; Gamage et al. 2023) and should be increased to at least 25% of the EU’s agricultural land by 2030 (EU’s Green Deal, Farm to Fork Strategy). Consumers are showing an increasing demand for reliable products, mainly from organic production (Mie et al. 2017; Wang et al. 2019). The main aspect of choosing organic products is buying and eating healthier foods (Rizzo et al. 2020; Hamilton and Hekmat 2018). In organic wheat production, crop stability and end-use quality are of great importance (Cesevičienė et al. 2009). Selection of varieties that produce high and stable yields even under marginal conditions are essential for both organic wheat growers and wheat breeders. Various studies were carried out with bread wheat to compare genotypes’ performances under organic and conventional management and dissect the genotype × management interaction (Campion et al. 2014; Mikó et al. 2014; Kissing Kucek et al. 2019; Herrera et al. 2020), whereas limited studies are available for durum wheat (Mikó et al. 2017).

The objectives of this study were (1) to examine the performance of a diverse set of durum wheat genotypes under different Hungarian growing conditions, (2) to detect useful varieties that could be recommended for organic farmers and (3) to identify phenotypic parameters and yield components for durum wheat breeding programmemes targeting low-input and organic farming systems.

Materials and methods

Plant material, field experiments and weather conditions

Thirty winter and spring durum wheat (Triticum turgidum L. subsp. durum (Desf.) Husn.) varieties and breeding lines were examined in the present study. The varieties originated from six countries (Table 1).

Table 1 Growth habit, year of release and origin of 30 durum wheat genotypes examined in Hungary between 2020 and 2022 in different management systems
Full size table

The durum wheat germplasm was evaluated between 2020 and 2022 in a randomised complete block design with three replications grown under low-input (LI), common conventional (CO), and officially registered organic (OR) management systems in Martonvásár (47°3′N, 18°8′E), Hungary. The conventional treatment used the same agronomic practises and fertilisation regime as the neighbouring farmers, the low-input treatment used 50% less fertilisation, whilst the organic treatment was on an organic certified field following organic practises. The soil type in the experimental fields was a chernozem with forest residues and its shallow layer did not contain lime and harmful salts. This layer had a neutral pH (pH = 7.0/6.7/7.2; LI/CO/OR) and, in terms of its physical properties, it was loam. Based on its humus content (2.4/3.0/2.7 m/m%), it had a moderate nitrogen supply, the phosphorus content was medium in LI and CO and good in OR (114/134/314 mg kg−1), the potassium supply was good (314/295/443 mg kg−1). In terms of microelements, zinc content of the soil was less than optimal with the exception of OR (1.1/1.3/3.4 mg kg−1), whilst copper (2.7/3.5/3.2 mg kg−1) and manganese (156/208/186 mg kg−1) contents were sufficient. The 6 m2 trial plots (row distance 0.15 m) were machine-drilled (HEGE-80 drill; Hans-Ulrich Hege GmbH & Co., Waldenburg, Germany) at optimal fall sowing date, except the 2020 organic trial which was sown late (12 November) due to unfavourable weather conditions (Suppl. Table S1). The experiments were combine-harvested (Wintersteiger Nursery Master Elite; Wintersteiger AG, Ried, Austria) at full maturity. No fungicides were applied in any trial; however, conventional and low-input plots were treated with herbicide and insecticide whenever it was needed.

During the three growing seasons, extreme weather conditions were observed, which were different from the thirty-year average. All three growing seasons were characterised by a mild winter, temperatures below − 10 °C were observed only three times. In all three years, the mean temperature between March and May was often below, whilst in June it was above than the thirty-year average (Suppl. Fig. S1). Precipitation was different both in distribution and in quantity during the three growing seasons (Suppl. Fig. S2). In November and December 2019, precipitation exceeded the 30-year average. In the following period, however, the amount of precipitation in most months was below the long-term average. More than half of the precipitation in June 2020 (45 mm) fell in one day, accompanied by strong windstorm. In the 2020–2021 growing season, October, February and May (and also July) were abundant in precipitation. In the following year, during the entire growing season, only April received a higher amount of precipitation compared to the 30-year average, but the entire year's precipitation was far below the long-term average.

Assessment of agronomic traits and technological quality of grains

Early spring ground cover was measured by the evaluation of the percentage of green canopy cover in 2021 and 2022 using the Canopeo® application (Patrignani and Ochsner 2015). Heading date, plant height and lodging were recorded before harvest. Resistance to powdery mildew and leaf spot diseases were evaluated according to Saari and Prescott (1975). In case of evaluation of leaf rust and stem rust the modified Cobb scale was used (Peterson et al. 1948). After combine harvesting the plots, grain yield was determined and calculated into t ha−1. Test weight (TW) was measured with a Infratec 1241 device (Foss Tecator AB, Höganäs, Sweden). The physical parameters of the grains, grain size (width, length), and thousand grain weight (TGW) were determined with a Marvin Digital Seed Analyser (Marvitech GmbH, Wittenburg, Germany). Measurements were carried out for each field replication.

Statistical analysis

Statistical analyses were performed using SPSS 16.0 software (SPSS Inc., Chicago, USA). Variance components and the significance level of main and interaction effects were calculated by general linear model (Univariate Analysis of Variance module). Canonical discriminant analysis was used to examine whether the 30 winter durum wheat genotypes in the three management systems and three years could be distinguished based on agronomic traits and grain physical parameters. Interactions and correlations between the traits and managements were determined using principal component analysis, whilst the best-performing genotypes were identified by the GGE Biplot Analysis module of Genstat 23rd ed. software (VSN International Ltd., Hemel Hempstead, UK). The strength of correlations was determined according to Evans (1996), who suggested five groups based on the absolute value of the correlation coefficient r: very weak (0.00–0.19), weak (0.20–0.39), moderate (0.40–0.59), strong (0.60–0.79) and very strong (0.80–1.00). Finally, a linear regression was used to describe the relationship more accurately between significantly correlated data. Repeatabilities (‘heritabilities’; h2) were calculated based on variance components according to Melchinger et al. (1998).

Results

In the present study ten traits were assessed on 30 durum wheat genotypes between 2020 and 2022 in three different management systems low-input (LI), common conventional (CO), and organic (OR). Our durum wheat breeding programme is mainly carried out under low-input conditions, and additionally the lodging and resistance of the durum wheat genotypes are used to be tested also under conventional conditions. According to our observations, the occurrence of the pathogens is more emphasised in CO having better nutrient supply than LI. Therefore, trial sites were analysed along this study in the order of importance within our current durum breeding system. Based on the ANOVA, both main factors (year, management, genotype) as well as their interactions were significant almost all evaluated traits (Table 2). In case of ground cover, the management × genotype interaction (M × G) and the three-way interaction were not significant. Significant M × G interactions indicate the different genotypic response to different management systems. The mean grain yield ranged between 5.81 and 6.02 t ha−1 in the different years, and between 5.15 and 6.57 t ha−1 in the different management systems. The highest mean yield was observed in 2021 in the conventional trial (7.86 t ha−1), whilst the lowest mean yield was realised in the same year in the organic trial (3.77 t ha−1). Across the three years, the mean yields in the common conventional, low input and organic trials were 6.57, 6.1 and 5.15 t ha−1, respectively (Suppl. Table S2). Repeatability values were low for for early spring ground cover (0.14), powdery mildew susceptibility (0.36) and lodging score (0.50), medium for grain yield (0.62), test weight (0.67), thousand grain weight (0.76), and high (0.82–0.99) for the other traits (Suppl. Table S3). Although the genetic determination of heading time was very strong, in our experiment the heading time was significantly influenced by the year. whilst 2020 was an average year in terms of heading date, the cold spring in 2021 and the early drought in the following year affected this trait. For example, in 2020 and 2022 the latest variety, ‘Mv Makaróni’ finished heading on May 21, whilst in 2021, due to the rainy and cold spring (April and May), 73% of the varieties started heading not before this date (Fig. 1a).

Table 2 Mean squares and significance levels of main and interaction effects of phenotypic parameters and yield components of 30 durum genotypes tested in Martonvásár, Hungary (2020–2022). All effects are highly significant at p < 0.001 unless otherwise indicated (**, p < 0.01; n.s., not significant)
Full size table
Fig. 1
figure 1

Trait variability in the durum diversity panel: a heading date depending on year; b disease incidence depending on year and management system (LI, low input; CO, common conventional; OR, organic); Martonvásár, Hungary (2020–2022)

Full size image

Mean plant height was highest under conventional management in 2021 (Suppl. Table S2). Lodging was observed mostly on the common conventional site where nitrogen fertilisation was highest. Test weight was between 78.78 kg hL−1 and 81.81 kg hL−1 with highest values reached under organic conditions. Similarly, thousand grain weight was significantly higher under organic management (48.72 g) compared to the low-input and common conventional system (42.90 g). Considering years, the highest grain weight was realised in 2021 (48.42 g), whilst the other two years were similar. Grain characteristics, width and length, were very stable across years and management systems and, therefore, had high repeatabilities (h2 > 0.85). Under organic conditions the durum grains were slightly broader, whereas slightly longer grains were observed under conventional management. Early spring ground cover was in 2021 in the conventional condition markedly different from that of the organic trial compared to the difference between management systems in the drier year 2022. Based on the means of the management systems, the organic field showed 27.2% and 34.2% less ground cover compared to the common conventional and the low-input trials (Suppl. Table S2).

Amongst naturally occurring pathogens, powdery mildew appeared every year and in every trial. In the wet year of 2021, leaf spot symptoms and traces of leaf and stem rust were also observed, especially under the conventional management with higher nitrogen input, where the pathogens appeared in more than 30% of the plots. Lodging in this trial may have had also on impact on disease incidence due to the wetter microclimate in the plots (Fig. 1b).

Based on the discriminant analysis, where all the phenotypic parameters and yield components were examined together, both the management systems (Fig. 2a) and the years (Fig. 2b) were found to be well separated from each other. With respect to the management systems, 78.5% of the low-input, 83% of the conventional and 81.1% of the organic observations were grouped correctly. The low input group was located exactly between the conventional and the organic groups (Fig. 2a). Considering years, 93.7% of the 2020 samples, 96.3% from 2021 and 95.9% from 2022 were placed in the appropriate group. The extremely dry year 2022 formed a transition between the years 2020 and 2021.

Fig. 2
figure 2

Classification of discriminant analysis based on all evaluated traits of the durum wheat diversity panel in Martonvásár, Hungary: a grouping according to management systems; b grouping accoriding to test years

Full size image

A more detailed analysis of the grain yield data was conducted using GGE biplot analysis. In the first analysis, including all trials, the first two principal components accounted for 69.7% and 12.1% of the GGE sum of squares, respectively, explaining a total of 82.0% of the variation (Fig. 3a). From the GGE biplot the significant genotype-by-management-by year interaction (G × M × Y) is clearly visible. whilst the low input trials show the lowest variability according to the angle between LI21 and LI20, variability is especially high within the organic trials. Comparison plots were created to assess the genotypes’ yield performance and specific adaptation to the three management systems. An ʻideal’ genotype with high grain yield across years is located in the innermost circle and nearby the arrow head of the ʻaverageʼ environment. For the low-input trials, this ideal genotype is’Miradoux’, followed by ‘Sambadur’ (Fig. 3b). In the conventional trials ʻMv Vékadur’ and ʻNS Žad’, followed by ‘Mv Masnidur’ are the best genotypes (Fig. 3c), whilst in the organic trials ʻSambadur’ lies exactly in the innermost circle, followed by ʻNS Žad’, ʻNS Dur’ and ʻMv Vékadur’ (Fig. 3d). Considering the test years, 2021 and 2022 show a similar response in the low-input (Fig. 3b) and organic trials (Fig. 3d), whereas in the conventioal trials these two years showed an opposite response (Fig. 3c). The old landrace selection ‘Cappelli’ and the Mediterranean germplasm was generally less adapted to the Hungarian test site.

Fig. 3
figure 3

GGE biplot analysis of durum wheat: a scatter plot of genotypes and enviornments (management × year); b comparison plot of low-input trials; c comparison plot of conventional trials; d comparison plot of organic trials. Managements: CO, conventional; LI, low input; OR, organic; Genotypes: AGH, Aghram; AZE, Azeghar 2; CAP, Cappelli; DOL, Dolap; DUR, NS Dur; FUE, Fuego; GAM, Gammary; GIB, Gibraltar; HFN, HFN96N; ICA, Icajin; IRI, Iride; LEV, Levante; LUN, Lunadur; M12, MVTD12-23; M16, MVTD16-19; M20, MVTD20-19; MAA, Maamouri 1; MAG, Mv Magnadur; MAK, Mv Makaróni; MAS, Mv Masnidur; MIR, Miradoux; OUS, Ouasloukos 1; PEL, Mv Pelsodur; SAM, Sambadur; SAR, Saragolla; SEB, Sebatel; SIM, Simeto; VEK, Mv Vékadur; VUL, Vulci; ZAD, NS Žad. Martonvásár, Hungary (2020–2022)

Full size image

The GGE biplot analysis allows also to identify the mean performance and stability of a genotype by a ranking and ‘which won where’ plot. From the ranking plot (Fig. 4a) it is obvious that four varieties, ʻNS Žad’, ‘NS Dur’, ‘Sambadur’ and ‘Mv Vékadur’, realised the highest grain yield across all three management systems. The distance to the abscissa, which represents the average environment, however, is significantly higher for ‘Sambadur’ and ʻNS Žad’ which indicates greater variability and, therefore, less stable performance, whereas ‘Mv Vékadur’ and ‘NS Dur’ show high and stable grain yields. ʻMv Pelsodur’ and ʻMv Magnadur’ are located exactly on the abscissa, indicating an extremely stable performance across management systems combined with an acceptable high yield potential. The ‘which-won-where’ plot (Fig. 4b) shows generally the same results, however, in another pattern by drawing sectors and thereby identifying environments with similar response as ‘mega-environments’. The two conventional systems, low-input (LI) and common conventional (CO), form together one mega-environment which is clearly separated from the organic one. The most extreme genotypes in the two sectors and, thus, best performing and adapted ones, are ‘Mv Vékadur’ and ‘NS Dur’ for the conventional systems, and ‘Sambadur’, ‘NS Žad’ and, with somewhat lower grain yield, breeding line MVTD12-23.

Fig. 4
figure 4

Mean performance and stability of 30 durum wheat genotypes tested under three different management systems: a ranking plot; b which won where plot including identification of mega-environments. Martonvásár, Hungary (2020–2022) Abbreviations for genotypes and management systems see Fig. 3

Full size image

Pearson’s correlation analysis were performed between the recorded traits (Fig. 5). Significant positive and high correlations (r > 0.8; p ≤ 0.01) were found between grain width and TGW. Early spring ground cover was measured in only two years and was positively correlated (r = 0.5) with grain yield (Fig. 5b). Linear regression analysis highlighted the positive relationship between grain width and TGW (R2 = 0.70), and that was even stronger when grain length was additionally considered in a multiple regression approach (R2 = 0.76) or if only data from the last two test years were considered (R2 = 0.862). Furthermore, early spring ground cover could explain 24.5% of the grain yield (Table 3).

Fig. 5
figure 5

Pearson’s correlation plot of phenotypic data (TW, test weight; heading date; lodging score; PH, plant height; PM, powdery mildew score; GL, grain length; GW, grain width; TGW, thousand grain weight; grain yield and early spring ground cover determined by Canopeo App) of 30 durum wheat genotypes evaluated in three management systems (low-input, conventional and organic): a 2020–2022; b 2021–2022. Critical r values for the correlation coefficients: 0.3494 (p < 0.05); 0.4487 (p < 0.01); 0.5541 (p < 0.001). Martonvásár, Hungary

Full size image
Table 3 Linear regression analysis using data of various phenotypic traits evaluated on 30 durum wheat genotypes grown in three management systems from 2020 to 2022 in Martonvásár, Hungary
Full size table

Discussion

In the next decades, radical changes are expected in field crop production. Within the framework of the ‘Green Deal’, the European Union announced its ‘Farm to Fork’ strategy (European Council 2024), which suggests actions for a more sustainable agriculture such as a decrease of the amount of pesticides and fertilisers applied in the fields by 2030. As organic farming is one of the possible solutions towards sustainability, EU member states were encouraged to develop national organic farming plans, to boost organic production to reach 25% of the EU’s agricultural land use by 2030. Although acreage under organic farming is growing and organic breeding programmemes are becoming more widespread, organic production is still mainly based on crop varieties bred for the conventional high-input sector. However, varieties suitable for organic farming need partly different phenotypic traits, e.g. nitrogen use efficiency, weed suppression or resistance to biotic and abiotic stresses. In the absence of these properties, 21% less yield can be achieved on organic land on average than under conventional farming conditions (Mikó et al. 2017; Pandino et al. 2020). Our results show 18.4% and 27.5% higher mean grain yields of durum wheat in the low-input trials compared to the conventional and organic trials, respectively. Comparing the minimum yields of each trial, a similar trend was observed, but the maximum grain yield was found in an organic trial. The genotypes that showed higher yield under organic conditions are probably better adapted to the low soil nitrogen availability. The study of Mariem et al. (2020) revealed that higher nitrogen fertilisation had a positive effect on average grain size, whilst it had no effect on TGW. Some genotypes responded to a lower nutrient level with higher grain yield, showing an outstanding nutrient use efficiency. Ingraffia et al. (2022) compared grain yields of durum varieties under organic conditions using normal tillage and no tillage. No tillage cultivation significantly reduced yield and protein content. The yield decrease was mainly due to a significant increase in weed biomass under this management condition. Similarly, we experienced a significant weed infestation at the organic trial site over the three years (data not shown), which may affect grain yield. The results of assessment of pathogens in wetter growing season (2021) revealed that the rusts appeared more recent under conventional conditions, than under the other two management systems due to better soil/nutritional properties and wetter microclimate. There is evidence that favourable weather conditions (Rodríguez-Moreno et al. 2020; Ajilogba and Walker 2023) and higher nitrogen fertilisation rates can increase the incidence of some leaf diseases (Simon et al. 2003; Schierenbeck et al. 2019; Luo et al. 2021) or stem rust (Wilcoxson 1980) of wheat. In the present study, analysis of variance showed that the main factors year (Y), management (M), genotype (G) and their interactions (Y × M, Y × G,) were significant for all the ten evaluated traits and M × G was significant for nine traits. For cereals, several other research studies showed that most of the main effects such as year, environment, management, genotype or their interactions were significant for the same traits as examined in this study (Cesevičienė et al. 2009; Mikó et al. 2014, 2017; Vida et al. 2022a, b; Al-Sayaydeh et al. 2023). Regarding the traits, significant M × G interaction was found in all cases, showing the different genotypic effects in the different management systems. Our experiment confirmed that based on the high repeatability (h2) values, the genetic determination of heading time, plant height, grain width and grain length are very strong (Sinclair and Jamieson 2006; Bányai et al. 2021), however, precipitation and temperature can play a significant role in the expression of morphological parameters (Bányai et al. 2020, 2021; Sun et al. 2020) or quality (Hacini et al. 2022; Vida et al. 2022a, b). Based on our results, heading time was significantly influenced by the year, although the raking of the varieties was the same (r = 0.93–0.96; p < 0.001). Heading time was shifted earlier or later depending on the growing season. Early ground cover is especially important in organic farming, both in terms of competition against weeds and preserving soil moisture. This trait was measured in three management systems, over two years using the Canopeo application. Early spring ground cover was not significant for G × M, but significant (p ≤ 0.01) for Y × G, therefore, varieties behaved similarly regardless of the management system but differently with regard to the year. A low repeatability was found for this trait, suggesting that no effective selection for this trait can be performed in early generations. The significant positive correlation between early ground cover with grain yield is in agreement with Govindasamy et al. (2022). A precise estimation of canopy cover using photogrammetry software such as Canopeo can help in yield prediction, crop quality assessment, and phenotyping research. In terms of an ʻideal’ genotype, which combines high grain yield with high stability, ʻMv Vékadurʼ and ʻNS Durʼ proved to be the best varieties in low-input and conventional farming, whilst ʻSambadurʼ, ʻNS Žadʼ and ‘MVTD12-23’ were the most stable and best performing genotypes in the organic trials. All these genotypes were bred in eastern Austria, Hungary and northern Serbia, hence, for the same production zone, the Pannonian region. Based on the discriminant and GGE biplot analyses, the organic trial site appeared very different from the two conventional ones, whilst the low-input and common conventional sites appeared to be more similar. Hence, in agreement with previous studies (Wolfe et al. 2008; Löschenberger et al. 2008; Lammerts van Bueren et al. 2011; Mikó et al. 2014, 2017), our experiment revealed the importance of evaluating relevant and useful traits of varieties (agronomic traits and technological quality) under both conventional and organic farming conditions. In our study, no varieties were found to provide stability and high yield performance in all management systems, which is in line with the findings of Mikó et al. (2017), who suggested that selection for organic farming should be carried out under organic conditions of the target country. Based on regression analysis, a significant positive relationship was found between grain width and TGW; considering also grain length in a multiple regression even improved the regression model. For grain weight and grain size also high h2 values were observed which is consistent with previous findings (Schierenbeck et al. 2021; Al-Sayaydeh et al. 2023).

Conclusions

A diversity panel of durum wheat was examined under different Hungarian growing conditions in three different growing seasons. No variety turned out to provide stable and high grain yields in all three management systems. It is therefore recommended that breeding of durum wheat should be targeted to the specific region and growing environment, however, it can be done under one management system for highly heritable traits such as plant height. Due to the low heritability, early ground cover can be selected in later generations, and its determination can help to predict grain yield because of the positive and significant medium correlation between the two traits. Based on their high heritability, grain width and length can together effectively be selected in early generations using digital imaging of grains. These traits work for indirect selection for high TGW due to the strong correlation between grain weight and grain size.