Introduction

Oilseed rape (Brassica napus L. var. oleifera ssp. napus AACC), also known as rapeseed or canola, is an important oil crop worldwide (Hasan et al. 2006; UFOP 2019). Canola seed oil is considered a major source of edible oil as well as an optimal raw material for green fuel production. Rapeseed oil is also used in the pharmaceutical and chemical industries (Friedt and Snowdon 2009). In 2022, annual global canola grain production reached 87 million tons; 19.41 million tons of this was in the European Union (EU27), with 4.51, 4.29 and 3.4 million tons from France, Germany and Poland, respectively. The average yield for winter oilseed rape (WOSR) in Poland ranged from 24.9 dt ha−1 in 2018 to 25.9 dt ha−1 in 2019 and to 32.3 dt ha−1 in 2022 (FAOSTAT 2024).

To feed the growing human population, further increase in canola grain production is needed. Development of WOSR breeding lines is an important goal of rapeseed breeding programs worldwide exhibiting both good oil quality, high agronomic value with satisfactory resistance to biotic and abiotic stresses (Würschum et al. 2012; Chen et al. 2014; Werner et al. 2018). Genetic information on the gene expression of quantitative traits in the early stage of the plant breeding process can help breeders to create an effective strategy for the development of new high yielding cultivars of canola. Many traits of agronomic importance are controlled by multiple gene of small effect, while their phenotypic expression is strongly affected by no-genetic factors as well as genotype by environment interactions (Piepho 1994; Nowosad et al. 2016; Werner et al. 2018; Bocianowski et al. 2019a). Therefore, increasing seed yield (SY) and SY components are the key goal of many studies and canola breeding programs (Würschum et al. 2012; Chen et al. 2014; Werner et al. 2018).

SY and yield-associated traits in crops are difficult to investigate because they are easily affected by changing environmental conditions. The genetic basis of these complex traits has been explored in many experiments on a range of crop plants, including rice (Rahimsoroush et al. 2021), barley (Wang et al. 2019), maize (Pan et al. 2017) and oilseed rape (Chen et al. 2014; Yang et al. 2017). It has been reported that most agronomical traits exhibit quantitative variation, which is controlled by multiple genes and is environmentally dependent. Moreover, Quarrie et al. (2006) pointed out that crop yield reflects the interaction of the environment with all growth and development processes that occur throughout the life cycle. One of the most important processes involved in crop acclimatization, adaptation and productivity is flowering time (FT). This developmental trait is determined by multiple genes that cause variability in the performance of genotypes and diverse responses to the cultivation environment, where important factors include cold exposure, season, day length, temperature, rainfall, and also biotic and abiotic stresses (Lou et al. 2007; Iniguez-Luy and Federico 2011; Fletcher et al. 2015; Raman et al. 2019). Long et al. (2007) demonstrated significant effects of genotype (G), environment (E) and GE interaction (GEI) on the FT of two segregating F2 populations comprising 202 and 404 lines of oilseed rape. Schiessl et al. (2015) showed that day-length and climatic conditions had an impact on the FT of 158 inbred lines of European winter-type B. napus. Moreover, high oil content in different genotypes was correlated with early flowering and a shorter flowering period.

Leon and Becker (1995) pointed out that every developmental process is, to a greater or lesser extent, under genetic control, and also dependent on environmental factors. SY is a typical quantitative trait in this respect, as it is influenced by many genetic factors as well as by interaction with the environment; consequently, particular genotypes can react differently to changing environmental conditions during plant development (Riaz et al. 2001; Zhang et al. 2013). Yan (2001) reported that SY of a certain genotype in a specific environment is influenced by G main effects, E main effects, and GEI. The authors demonstrated that, in field trials carried out under different agroecological conditions, 80% of yield variation was usually attributable to E, while G and GEI each caused 10% of the variation. Similarly to SY, the expression of SY-related traits, such as the number of branches, the number of siliques per plant, and the number of seeds per silique resulted from the G × year (environment) interactions (Chen et al. 2014; Bocianowski et al. 2017; Łopatyńska et al. 2021). Moreover, changes in climate may significantly influence crop growth and yield, and an understanding of the effects of environmental factors on crop development and growth can improve the selection of specific strains for cultivation in the target regions (Marjanović-Jeromela et al. 2011; Zhang et al. 2013; Bocianowski and Liersch 2021). In particular, comprehensive analyses of SY and yield-associated traits, in addition to their interaction with environmental factors, are crucial for rapeseed yield improvement (Shi et al. 2009; Iniguez-Lay and Federico 2011; Zhang et al. 2013). In addition, Chen et al. (2021) investigated transient daily heat stress during the early reproductive phase of canola under controlled environment. The authors showed that heat stress treatments during the week before first open flower on the main stem or during the first week or second week following first open flower greatly reduced pod number and seed yield. They also defined the conditions for controlled environment screening for heat tolerance in B. napus.

In many crops species the multi-environmental analysis [genotype main effect (G), environment (E) and genotype by environment interaction (GE)] of different breeding materials is widely used for characterization of genotypes from an agronomic point of view. Many researchers tried to define the genotypes suitable for various environments with high adaptability and good stability. For this reason, plant breeders often conduct factorial field experiments to study the influence of GEI on crop yield and yield components. This influence may be assessed by combination analysis of variance (ANOVA), for the G and E main effects, and principal component analysis (PCA), including multiplicative parameters, for GEI effects. The AMMI biplot graphic simultaneously displays both main and interaction effects for genotypes and environments, and enables a single analysis of GEI. Therefore, the AMMI model is also known as ‘interaction PCA’ (1990). This model has been used to analyse GEI in field trials of many crops, such as maize (Branković-Radojčić et al. 2018), wheat (Kumar et al. 2018; Singh et al. 2019), soybean (Sneller and Dombek 1995), ethiopian mustard (Tadesse et al. 2018), mustard and canola (Gunasekera et al. 2006; Alizadeh et al. 2020; Zulfqar et al. 2021; Qasemi et al. 2022; Arinaitwe et al. 2023; Okla et al. 2023; Shojaei et al. 2023; Oghan et al. 2024).

The use of the AMMI model is a tool for identifying the genotypes best suited to specific soil and climatic conditions. Measures of stability allow to distinguish stable and specific genotypes. One of the best measures of genotype stability is the AMMI stability value coefficient based on the first two interacting principal components (Bocianowski et al. 2019b, 2021; Bocianowski and Prażak 2022).

The main objective of this study was to use the AMMI model to estimate GEI for 25 WOSR genotypes in six environments for the following SY-related traits: beginning of flowering (BF), SY and SY components i.e. the number of siliques per plant (NSP), the length of siliques (LS), the number of seeds per silique (NSS), and the weight of 1000 seeds (WTS).

Materials and methods

Plant material and field experiment

A collection of 25 genotypes of WOSR (Brassica napus L.) was used in this study: seven Polish and European cultivars (Monolit, Brendy, Starter, Mendel, Polka, Sherlock, Zornyj), three CMS ogu-INRA hybrids (F1_952, F1_S2, F1_239), six parental forms of the ogu-INRA system, i.e. three CMS lines (CMS_64, CMS_313, CMS_1612) and three DH restorer lines (Rfo_37, Rfo_38, Rfo_39), DH breeding line (DH D × C), double low quality breeding lines (HO_TP, HO_TP_00, 00_SS), a yellow-seeded line (Z 114), a high-oleic mutant line (HO_SS), and a mutant recombinant line with changed unsaturated fatty acid composition in seed oil (i.e., high oleic and low linolenic) (HOLL_SS), as well as a semi-resynthetic (semi-RS) WOSR lines (R43, S1) (Table 1). Hybrid seeds resulting from crosses between the male-sterile CMS lines and the restorer lines has been produced in isolation tents under field conditions. Field trials were performed in four replicates of a randomized block design at the Experimental Station of Plant Breeding Company Strzelce Ltd. in Borowo (E1, E2 and E3, respectively) and Plant Breeding Company Smolice Ltd. in Łagiewniki (E4, E5 and E6, respectively), Poland, during the crop seasons 2014/2015, 2015/2016 and 2016/2017. Genotypes were grown in a four-row plots of 10 m2 (Borowo) and 9.6 m2 (Łagiewniki) with a 0.30 row distance and a sowing density of 70 seeds m−2. Agricultural practice were optimal for local agroecological conditions in both of the investigated locations. Field experiments were conducted on podzolic soil of quality class IIIa (E1, E2) and IVa (E3) in Borowo and on typical brown soil of quality class IIIa in Łagiewniki (E4–E6). The more details of the plant materials and the field trials were described by Liersch et al. (2020). The monthly temperatures and precipitations in each growing season, from the end of August (sowing) until the end of July of the next year (harvest), are presented in Fig. 1.

Table 1 Genotype description
Full size table
Fig. 1
figure 1

Weather conditions during growing seasons in six environments (E1–E6); dark grey bars, mean monthly precipitation in growing season [mm]; grey bars many-years monthly precipitation [mm]; black solid and grey dashed lines, mean monthly temperatures in growing season and mean many-years monthly temperatures [°C], respectively

Full size image

Quantitative traits

Experimental data were collected on both morphological and seed features. The beginning of flowering time (days) was defined as the number of the days from the beginning of the year to the beginning of flowering (i.e. when 50% of plants have their first flowers). The number of siliques per plant was the number of well-filled, normally developed siliques from five, randomly selected plants from the middle of each plot. At the maturity stage, siliques from the primary branches of five representative plants from the center of each plot were collected and then dried. Silique length (mm) and the number of seeds per silique were determined on twenty-five well-developed siliques. Seeds were harvested at the seed maturity stage using a plot harvester. Seed yields per plot, which was either 10.0 m2 (Borowo) or 9.6 m2 (Łagiewniki) in area, were measured and converted to dt∙ha−1 for all statistical analyses. The weight of a 1000 seeds (g) was calculated from the average of three measurements from the mixed seeds of each plot.

Statistical analysis

Data were analyzed using the additive main effects and multiplicative interaction model Gauch and Zobel (1990) for each trait, independently. The AMMI model first fits additive effects for the main effects of genotype (G) and environment (E) followed by multiplicative effects for GEI by PCA. The results of AMMI analysis are presented in biplot graphs. The AMMI model (Nowosad et al. 2016) is given by:

$$y_{ge} = \mu + \alpha_g + \beta_e + \sum_{n = 1}^N {\lambda_n \gamma_{gn} \delta_{en} } + Q_{ge} ,$$

where yge is the trait mean of genotype g in environment e, μ is the grand mean, αg are the genotypic mean deviations, βe are the environment mean deviations, N is the number of PCA axis retained in the adjusted model, λn is the eigenvalue of the PCA axis n, γgn is the genotype score for PCA axis n, δen is the score eigenvectors for PCA axis n, Qge is the residuals, including AMMI noise and pooled experimental error. The AMMI stability value (ASV) was used to compare the stability of genotype as described by Purchase et al. (2000):

$$ASV = \sqrt {{\left[ {\frac{{SS_{IPCA1} }}{{SS_{IPCA2} }}\left( {IPCA_1 } \right)} \right]^2 + \left( {IPCA_2 } \right)^2 }} ,$$

where SSIPCA1 is the sum of squares for IPCA1, SSIPCA2 is the sum of squares for IPCA2, and the IPCA1 and IPCA2 scores are the genotypic scores in the AMMI model. A lower ASV score indicates a more stable genotype over several environments (Nowosad et al. 2017).

The genotype selection index (GSI) was calculated for each genotype. This incorporates both mean of trait and ASV index in a single criterion (GSIi) as described by Farshadfar and Sutka (2003):

$${\text{GSI}}_i = {\text{RM}}_i + {\text{RA}}_i ,$$

where RMi is the rank of trait mean (from maximal to minimal for SY the NSP, LS, NSS and WTS, and from minimal to maximal for BF) for the i-th genotype, and RAi is the rank of the ASV for the i-th genotype. Finally, the total genotype selection index (TGSI) was calculated for each genotype as the sum of GSIs for all six traits in the study. Correlation analysis between observed traits was performed on genotypic averages. All analyses were conducted using the GenStat v. 18 statistical software package.

Results

We analyzed 25 OSR genotypes including open-pollinated and hybrid cultivars, breeding lines of canola type, Ogura CMS and Rfo canola breeding lines and lines with different C18 mono- and polyunsaturated fatty acid content in seed oil and lines characterized by different seed colour. Mean values of all traits are shown in Table 2.

Table 2 Mean values of seed traits of 25 genotypes tested in field trials conducted in six environments
Full size table

The three sources of variation (G, E and GEI) were highly significant for all six observed traits of WOSR (Table 2).

Beginning of flowering (BF)

The impact of the environment on BF was relatively low, but significant, representing 10.80% of the total variation. The effects of G explained 67.16% of the variation, while the effects of GEI explained 19.50% of the variation (Table 3). Values for the three principal components were also found to be highly significant and jointly accounted for 92.59% of overall variation in BF. IPCA 1 accounted for 46.75% of the variation caused by the interaction, while IPCA 2 and IPCA 3 accounted for 33.66% and 12.18% of the variation, respectively. Across the 25 genotypes in the six environments, BF varied from 109.5 days for Starter (G03) and Rfo_39 (G09) in E2 (Borowo 2016) to 130.8 days for HO_SS (G13) in E6 (Łagiewniki 2017), with the average BF of 118.5 days (Table 4). Across all environments, the cultivar Starter (G03) was the earliest to flower (the average of 112 days), while the latest was the genotype HO_SS (G13) (the average of 126 days). In each environment, the average BF across all genotypes varied from 117 in E5 (Łagiewniki 2016) to 121.1 days in E3 (Borowo 2017) (Table 4). The stability of the genotypes in the six environments with respect to BF was visualized as a biplot (Fig. 2). The WOSR genotypes interacted differently with respect to the climate conditions in each environment. The genotypes HOLL_SS (G15), S43 (G20) and cultivar Zornyj (G23) interacted positively with environment E5 (Łagiewniki 2016), but negatively with environments E1 (Borowo 2015) and E2 (Borowo 2016) (Fig. 2). The cultivars Starter (G03), Mendel (G04) and Rfo_39 (G09) interacted positively with E4 (Łagiewniki 2015) and E6 (Łagiewniki 2017), but negatively with E2 (Borowo 2016). The analysis showed that some genotypes have a high adaptation capacity; however, most of them have specific adaptability. AMMI stability values (ASVs) revealed variations in BF stability among the genotypes (Table 4). The Z_114 (G05), Brendy (G02) and F1_239 (G25) genotypes, which showed an ASV of 0.198, 0.393 and 0.397, respectively, were found to be the most stable, while HO_SS (G13) and cultivar Zornyj (G23) with an ASV of 3.358 and 2.311, respectively, were the least stable (Table 4). Genotype G05, with a low average BF (115.7 days), had the best ASV (0.198) and genotype selection index (GSI) (6), while the HO_SS (G13) breeding line was characterized by the worst GSI (50).

Table 3 Mean squares (m.s.) from analysis of variance of main effects and interactions for six observed traits in winter oilseed rape genotypes, with variability explained (ve, in %)
Full size table
Table 4 Average beginning of flowering (days) for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table
Fig. 2
figure 2

Biplot for genotype by environment interaction of beginning of flowering in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image

Seed yield (SY)

ANOVA showed that the sum of squares for the environment main effects represented 42.15% of the SY variation. The differences between the genotypes explained 36.00% of the SY variation, while the effects of GEI explained 11.52% (Table 3). The values of the three principal components were also found to be highly significant and jointly accounted for 81.61% of the total variation. IPCA 1 accounted for 41.29% of the variation caused by the interaction, while IPCA 2 and IPCA 3 accounted for 22.48 and 17.84%, respectively. SY of the 25 genotypes across the six environments varied from 5.5 dt ha−1 for HO_SS (G13) in Borowo 2017 (E3) to 61.33 dt ha−1 for Starter (G03) in Łagiewniki 2015 (E4), with an average of 29.95 dt ha−1 (Table 5). The cultivar Starter (G03) had the highest average SY (41.08 dt ha−1), while the genotype HO_SS (G13) had the lowest (10.5 dt ha−1). The average SY in each environment varied from 16.55 dt ha−1 in E2 to 41.64 dt ha−1 in E4. The genotypes HO_SS (G13), HOLL_SS (G15), HO_TP (G17), as well as cultivar Zornyj (G23), interacted positively with environment E2 (Borowo 2016), but negatively with E1 (Borowo 2015) and E5 (Łagiewniki 2016) (Fig. 3). Cultivar Mendel (G04) and the S1 breeding line (G21) interacted positively with E4 (Łagiewniki 2015), but negatively with E3 (Borowo 2017) and E6 (Łagiewniki 2017). Genotypes DH_D × C (G05) and CMS_64 (G10) interacted positively with E5 (Łagiewniki 2016). The HO-type cultivar Polka (G16) and the CMS_313 line (G11) returned ASVs of 0.351 and 0.447, respectively, and were thus relatively stable, while genotypes HO_SS (G13) and HOLL_SS (G15), with ASVs of 5.817 and 4.223, respectively, were the least stable (Table 5). Cultivar Brendy (G02) and genotype CMS_313 (G11), with average SYs of, respectively, 36.6 and 33.12 dt ha−1, and ASVs of, respectively, 0.866 and 0.447, were characterized by the best GSI value of 11, compared to the HO_SS (G13) line, which had the worst GSI value (50).

Table 5 Average seed yield (dt ha−1) for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table
Fig. 3
figure 3

Biplot for genotype by environment interaction of seed yield in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image

Number of siliques per plant (NSP)

The three sources of variation (G, E, GEI) were found to be highly significant with respect to NSP. The impact of the environment on this trait was low, albeit significant, representing only 6.26% of the variation. The differences between genotypes explained 57.25% of the variation in NSP, while the effects of GEI explained 24.94% (Table 3). The values of the three principal components (PCs) were also found to be highly significant and jointly accounted for 87.92% of overall variation in NSP. IPCA 1 accounted for 51.51% of the variation caused by the interaction, while IPCA 2 and IPCA 3 accounted for 23.86 and 12.55% of the variation, respectively. Among the tested genotypes, NSP varied across the six environments from 138.4 for HO_SS (G13) in E3 (Borowo 2017) to 392 for F1_239 (G25) in E3 (Borowo 2017), with an average of 263.17 (Table 6). The hybrid F1_S2 (G24) had the highest average NSP (335.2), while genotype HO_SS (G13) had the lowest NSP (166.1). The average NSP in each environment varied from 246.3 in E6 to 283.4 in E1. The cultivar Monolit (G01), as well as the HOLL_SS (G15), S1 (G21) and F1_239 (G25) breeding lines interacted positively with environments E1 (Borowo 2015) and E3 (Borowo 2017), but negatively with E4 (Łagiewniki 2015) and E6 (Łagiewniki 2017) (Fig. 4). The Z114 (G06), CMS_1612 (G12) and HO_SS (G13) breeding lines and Zornyj (G23) interacted positively with E5 (Łagiewniki 2016), but negatively with E2 (Borowo 2016). The F1_952 hybrid (G19) and Rfo_38 (G08) returned an ASV of 0.892 and 1.034, respectively, showing them to be the most stable, while the S43 (G20) and F1_239 hybrid (G25) genotypes, with an ASV of 18.883 and 12.562, respectively, were the least stable (Table 6). The Rfo_38 (G08) breeding line with average NSP equal to 291.2 and an ASV of 1.034 had the best GSI (10), while genotype HO_SS (G13) was characterized by the worst GSI value (45).

Table 6 Average number of siliques per plant for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table
Fig. 4
figure 4

Biplot for genotype by environment interaction of the number of siliques per plant in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image

Length of siliques (LS)

The three sources of variation were highly significant with respect to LS. However, genotype was the most important factor. The environment main effect represented only 3.99% of the total variation. The differences between genotypes explained 57.04% of the total variation in LS, while the effects of GEI represented 28.44% of the variation (Table 3). The values of the first three PCs were also found to be highly significant and jointly accounted for 79.44% of the overall variation in LS. IPCA 1 explained 31.77% of the variation caused by the interaction, while IPCA 2 and IPCA 3 accounted for 28.29 and 19.37%, respectively. Across the 25 genotypes in the six environments, LS varied from 41.05 mm for the S43 strain (G20) in E1 (Borowo 2015) to 74.5 mm for Zornyj (G23) in E6 (Łagiewniki 2017), with an average of 61.59 mm (Table 7). Cultivar Zornyj (G23) had the highest average LS (69.29 mm), while the S43 (G20) genotype had the lowest (50.79 mm). The average LS in each environment varied from 60.02 mm for Łagiewniki 2015 (E4) to 63.72 mm for Łagiewniki 2016 (E5). The stability of the genotypes in the six environments with respect to LS was visualized as a biplot (Fig. 5). Cultivars Monolit (G01), Polka (G16) and S1 (G21) interacted positively with environment E3 (Borowo 2017), but negatively with E6 (Łagiewniki 2017) (Fig. 5). Cultivar Brendy (G02), together with HO_SS (G13) and HO_TP_00 (G18), interacted positively with environments E1 (Borowo 2015) and E4 (Łagiewniki 2015), but negatively with environments E2 (Borowo 2016) and E5 (Łagiewniki 2016). The Rfo_39 (G09) and Rfo_37 (G07) breeding lines, with an ASV of 0.263 and 0.388, respectively, were identified as the most stable, while the S43 breeding line (G20) and the F1_952 hybrid (G19), with an ASV index of 2.942 and 2.286, respectively, were the least stable (Table 7). The cultivar Sherlock (G22), with a high average LS (66.89 cm) and an ASV index of 0.417, had the best GSI of 7 (Table 7), while the S43 breeding line (G20) had the worst GSI (50).

Table 7 Average length of silique (mm) for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table
Fig. 5
figure 5

Biplot for genotype by environment interaction of length of silique in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image

Number of seeds per silique (NSS)

The three sources of variation were found to be highly significant with respect to NSS. In ANOVA, the sum of squares for the environment main effect represented 21.86% of the total variation in NSS, while the differences between the genotypes explained 38.15% of the total variation. At the same time, the effects of GEI explained 30.83% of the variation in NSS (Table 3). The values of the three principal components were also found to be highly significant and jointly accounted for 78.94% of the total variation in NSS. IPCA 1 accounted for 38.15% of the variation caused by the interaction, while IPCA 2 and IPCA 3 represented 25.03 and 15.75%, respectively. Of the 25 genotypes across the six environments, NSS varied from 10.1 for the Rfo_39 breeding line (G09) in E2 (Borowo 2016) to 26.02 for the Rfo_39 breeding line (G09) in E4 (Łagiewniki 2015), with an average of 18.24 (Table 8). The cultivar Sherlock (G22) had the highest average NSS (23.19), while the genotype HOLL_SS (G15) had the lowest (13.61). The average NSS in each environment varied from 15.88 in E2 to 21.24 in E4. The stability of the genotypes with respect to NSS was visualized as a biplot (Fig. 6). The genotypes Rfo_37 (G07), Rfo_39 (G09) and HO_SS (G13), as well as S1 (G21), interacted positively with environments E1 (Borowo 2015) and E4 (Łagiewniki 2015), but negatively with E2 (Borowo 2016), E3 (Borowo 2017) and E6 (Łagiewniki 2017) (Fig. 6). Cultivar Monolit (G01), as well as F1 hybrids G19 and G25, interacted positively with E2 (Borowo 2016), but negatively with E5 (Łagiewniki 2016). Cultivars Zornyj (G23) and Starter (G03), as well as the HO_TP (G17) breeding line, with ASVs of 0.075, 0.276 and 0.232, respectively, were found to be the most stable, while the genotypes CMS_1612 (G12) and Rfo_37 (G07) revealed the lowest stability, of 2.365 and 2.348, respectively (Table 8). Cultivar Mendel (G04) and the HO_TP (G17) line revealed a high average NSS (21.47 and 19.8, respectively) and high stability (0.822 and 0.232, respectively). They were also characterized by the best GSI values (8). The Rfo_37 (G07) line had the worst GSI (45).

Table 8 Average number of seeds per silique for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table
Fig. 6
figure 6

Biplot for genotype by environment interaction of the number of seeds per silique in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image

Weight of 1000 seeds (WTS)

ANOVA showed that the sum of squares for genotype main effects represented 48.26% of the total variation in WTS indicating that the trait depended mostly on G. The differences between the environments explained 27.19% of the total variation in WTS, while the effects of GEI explained 16.42% (Table 3). The values of the three PCs were also found to be highly significant and jointly accounted for 85.91% of the overall effect on variation in WTS. IPCA 1 accounted for 45.57% of the variation caused by the interaction, while IPCA 2 and IPCA 3 represented 24.11% and 16.23%, respectively (Table 3, Fig. 7). Of the 25 genotypes across the six environments, WTS varied from 3.33 g for S43 (G20) in E1 (Borowo 2015) to 6.357 g for the DH_D × C breeding line (G05) in E4 (Łagiewniki 2015), with an average of 5.005 g (Table 9). The F1 hybrid 239 (G25) had the highest average WTS (5.748 g), while the HO_SS breeding line (G13) had the lowest (4.048 g). The average WTS value in each environment varied from 4.63 g in E2 (Borowo 2016) to 5.344 g in E4 (Łagiewniki 2015). The stability of the genotypes in the six environments with respect to WTS was visualized as a biplot (Fig. 7). The breeding lines Rfo_38 (G08) and HOLL_SS (G15) interacted positively with E2 and E5, but negatively with E4 (Fig. 7). The genotypes G06, G07, G20 and G21 interacted positively with E3 (Borowo 2017), but negatively with E1 (Borowo 2015). The F1 hybrids 239 (G25) and 952 (G19), with an ASV of 0.021 and 0.092, respectively, were found to be the most stable, while DH_D × C (G05) and S43 (G20), with an ASV of 1.341 and 1.112, respectively were the least stable (Table 9). A F1 hybrid 239 (G25) with the highest average WTS (5.748 g) and an ASV of 0.021 also had the best GSI (2), whereas the S43 (G20) line had the worst GSI value (48).

Fig. 7
figure 7

Biplot for genotype by environment interaction of weight of 1000 seeds in winter oilseed rape genotypes in six environments, showing the effects of primary and secondary components (IPCA 1 and IPCA 2, respectively). E1—Borowo 2015, E2—Borowo 2016, E3—Borowo 2017, E4—Łagiewniki 2015, E5—Łagiewniki 2016, E6—Łagiewniki 2017

Full size image
Table 9 Average weight of 1000 seeds (g) for genotypes and environments, principal component analysis values, AMMI stability value (ASV), rank of the AMMI stability value (RA), rank of trait mean (RM, from minimal to maximal), and genotype selection index (GSI) of 25 winter oilseed rape genotypes
Full size table

Total genotype selection index (TGSI)

Jointly for the six analyzed traits, the best total genotype selection index value (96.5) was returned for the F1 hybrid 239 (G25) while the HO_SS (G13) line had the worst (260) (Table 2). Therefore, the G25 genotype can be recommended for inclusion in further breeding programs due to its stability and good average values of SY-related traits.

Correlation analysis

Seed yield was positively correlated with: the number of siliques per plant (0.52), the length of siliques (0.55), the number of seeds per silique (0.72) and the weight of 1000 seeds (0.57); while negatively correlated with beginning of flowering (−0.77). In addition, beginning of flowering was negatively correlated with: the number of siliques per plant (−0.43), the number of seeds per silique (−0.48) and the weight of 1000 seeds (−0.53). The length of siliques was positively correlated with the number of seeds per silique (0.69) and the weight of 1000 seeds (0.59) (Fig. 8).

Fig. 8
figure 8

Heat map of correlation coefficients between observed rapeseed traits

Full size image

Discussion

The yield of major crop plants is a consequence of plant development, which depends on an appropriate response to environmental and agrotechnical conditions, including adaptation of the plant reproductive system to the prevailing climatic conditions (i.e., vernalization, water supply, day length and temperature), the response to biotic and abiotic stresses, such as storms, in addition to timely and properly performed agrotechnical treatments. In our study we showed a significant influence of G, E and GEI on all rapeseed traits investigated: the beginning of flowering time, seed yield and seed -yield components, i.e., the number of siliques per plant, the length of siliques, the number of seeds per silique and the weight of 1000 seeds.

At both localities, Borowo and Łagiewniki, there was variability in soil type and water supply across the site, i.e., the field experiments were conducted on different types of soil. At Borowo, the various soil types were classified as light, loamy sand and as “medium” or “heavy soil”, while at Łagiewniki the soil was sandy clay loam. The topsoil was slightly acidic, and the pH values ranged from 6.0 (E1, E2, E3) to 6.2 (E4 and E6). Agricultural practices were optimal for the agroecological conditions, and the timing and rate of fertilization were the same in each environment. At the same time, the field trial results demonstrated the impact of weather, especially of temperature and rainfall, on plant development, including germination and wintering (not shown), flowering time, SY and SY-related traits.

Crop-plant species grown under different agroecological conditions reveal phenotypical differences that arise in response to particular environmental parameters, such as soil type, agrotechnical treatments, in addition to weather conditions, such as cold exposure, day length and rainfall. The beginning of flowering time and length of flowering time are determined by multiple genes and quantitative trait loci localized in both the A and C genomes of B. napus, thus leading to a diverse response to the cultivation environment among the various rapeseed genotypes (Long et al. 2007; Iniguez-Luz and Federico 2011; Raman et al. 2011; Fletcher et al. 2015; Schiessl et al. 2015; Ghanbari et al. 2020). Osborn et al. (1997) reported the flowering time as a trait with a high degree of heritability but showing a strong interaction with environmental conditions. In populations of 391 elite OSR and 156 WOSR genotypes, heritability (h2) was 0.91 and 0.84, respectively (Würschum et al. 2012), and ranged from 77 to 90% in other populations (Long et al. 2007). However, the results showed both G and G × E variance in these populations. Similar results were reported by Bocianowski et al. (2017; 2018) for different WOSR cultivars and breeding materials. In our experiment, BF varied considerably with genotype, year, and location, and depended on the temperature in April: BF ranged from 117 days (Łagiewniki 2016) to 121.1 days (Borowo 2017), and flowering began earlier when the average temperature in April was higher (Borowo 2015, 2016 and Łagiewniki 2016).

In this study, the G and GEI variance was significant for SY and yield-related traits of the 25 WOSR genotypes. In both locations, genotype had a significant impact on SY, NSP, LS, NSS and WTS. At the same time, we found a statistically significant influence of E and GEI on the traits analyzed. Studies of G, E and GEI for traits related to plant development and SY in OSR are frequent (Seyis et al. 2006; Spasibionek et al. 2020; Bocianowski and Liersch 2021). Additionally, many reports have pointed out that traits of agronomic importance are controlled by multiple genes, whose expression is the result of the interaction between genotype and environmental factors (Zhao et al. 2005; Long et al. 2007; Shi et al. 2009; Fletcher et al. 2015). Szała et al. (2019) observed a G and E main effect as well as GEI in field trials including fifteen F1 Ogura hybrids and their parental lines in three locations. Both Bocianowski et al. (2017) and Spasibionek et al. (2020) studied SY of different genotypes of WOSR and determined GEI as well the G effect based on the results of field experiments. Nowosad et al. (2016) showed that SY of 25 WOSR genotypes significantly depended on both genetic and environmental factors, where climate conditions were the most influential factors in their experiments in the western part of West Poland. In our study, we observed strong environmental factors, especially precipitation and soil water availability, which depended on the type of soil. A higher SY was observed in Łagiewniki (E4–E5)—respectively 41.64, 30.75 and 28.84 dt ha−1—due to its likely more fertile soil than found in Borowo (E1–E3), where SY was lower, respectively 36.13, 16.55 and 25.84 dt ha−1. Additionally, during the seed maturation period just before harvest, heavy rainfall in July negatively influenced SY in Borowo in 2016 (E2).

NSP and the number of plants per unit area are generally considered to be the most important traits affecting SY in B. napus, followed by NSS and WTS (Wójtowicz 2013). ANOVA indicated that the main effects of genotype and environment, as well as GE interaction, were significant for all yield component traits. Significant differences between the environments (two years) and agronomic traits of WOSR, such as WTS, NSP and SPS, were noted by Chen et al. (2014) and Wolko et al. (2019). ANOVA performed by Shi et al. (2011) revealed that G, E and GEI have significant effects on the performance of 15 yield-correlated traits in rapeseed. Also, Bocianowski et al. (2017), in their field trials of seven different types of WOSR cultivar, observed a significant effect of cultivar × location interaction for LS and BF, and significant effects of cultivar × year × location for the beginning, end and length of flowering, as well as for WTS and oil content. In our research, yield-correlated traits were primarily defined by genotype, and, to a lesser extent, by E and GEI. NSP was higher in Borowo (283.4, 273.2 and 271.6 for E1, E2 and E3, respectively) than in Łagiewniki (255.3, 249.2 and 246.3 for E4, E5 and E6, respectively). The values of other yield components were higher in Łagiewniki. The average LS in Borowo (E1-E3) was respectively 61.02, 60.31 and 62.76 mm, while in Łagiewniki (E4-E6) it was 60.02, 63.72 and 61.68 mm, respectively. The average NSS varied across the genotypes from 18.88 (E2) to 18.54 (E1) in Borowo, and from 17.48 (E6) to 21.24 (E4). At the same time, the mean value of WTS was higher in Łagiewniki (4.969–5.344 g) than in Borowo (4.63–5.141 g). Wójtowicz (2013) concluded in their study that environment has the strongest influence on NSP and the number of plants per unit area, rather than NSS and WTS. At the same time, the author pointed out that water shortage during the flowering period and during the formation of siliques was crucial in limiting the plants’ ability to produce siliques. We noted a similar relationship in our study. The number of generative organs per plant, the number and length of siliques, SY and seed-yield components depend strongly on the humidity conditions and availability of water during the growing season.

In agronomic research, use of the correct statistical model is crucial for revealing how the quantitative traits of the various genotypes studied are affected by environmental factors and agrotechnics (Schmidt et al. 2001; Ma et al. 2004). Environmental conditions may have a different effect on phenotypic traits depending on the particular genotype, thus leading to different GEI. The stability or instability of a genotype can be determined by a series of multi-environment field trials. When phenotypic traits, such as yield and its components, vary under different environmental conditions, G is assessed as unstable. From the breeders’ and farmers’ point of view, G or G with little E influence is the most desirable. In our study, we tested several canola type WOSR accessions with different genetic backgrounds: cultivars of different origin, F1 Ogura hybrids, CMS ogura and restorer lines, semi-resynthesized lines, and yellow-seeded lines, recombinant breeding lines, and lines with different fatty acid composition or breeding lines typically used for canola seed oil production. In plants, all life processes depend on temperature and availability of water, which is used for nutrient uptake and transport, transpiration, and all biochemical and physiological processes. The weather conditions in Poland varied over the three growing seasons in our study (Fig. 1, Fig. S1). In 2014/2015, there were strong abiotic stresses caused by drought, from December 2014 to the end of June 2015. The sum of rainfall in winter and spring 2014/2015 was 191.4 mm in Borowo and 195.8 mm in Łagiewniki, being markedly lower than the average for the period 1957–2016 (318.0 mm in Borowo; 326.0 mm in Łagiewniki, respectively). Also, in 2015/2016 the weather conditions were unfavorable; the lack of snow cover and extremely low temperatures at the beginning of January 2016 caused severe damage to the plants. Moreover, heavy rainfall in July 2016, just before harvest, significantly reduced the SY in Borowo (Fig. 1, Fig. S1). In this study we observed that the weather in addition to the soil type significantly influenced plant development which was reflected in the SY and SY-trait values. At the same time, many authors have shown that quantitative traits in WOSR are affected by the different responses of particular genotypes to varying environmental conditions during development (Shi et al. 2009; Fletcher et al. 2015; Thian et al. 2017; Spasibionek et al. 2020).

In this study, we applied the AMMI model for statistical analyses. This is a two-factor model describing the sum of the main additive effects for G (e.g., varieties, lines, soil cultivation methods), E (year, location or a combination of year and location), and the GEI in a multiplicative form. AMMI models for interpreting GEI have been used in several quality breeding programs and also in plant breeding multi-environmental trials (Vargas et al. 1999; Farshadfar and Sutka 2003; Nowosad et al. 2016). Using this approach, we observed significant GEI for BF and SY-related traits. The TGSI index values allowed us to assess the adaptability and stability of genotypes and to discriminate among the genotypes for particular selection and breeding strategies.

Analysis of genotype stability is a very important part of research and breeding work. Especially when there are large soil and climate variations. A large emphasis on stability studies in research was conducted by: Verma and Singh (2020), Jędzura et al. (2023), Oroian et al., (2023), Filio et al., (2023), Hossain et al. (2023), Kumar et al. (2023), Eradasappa et al. (2024), van der Merwe et al. (2024). In the research presented here, we distinguished genotypes stable through all the environments considered.

Conclusions

This study illustrates how multivariate statistical analysis can describe the interaction effect of a genotype with a particular environment for SY and other quantitative traits in WOSR. AMMI analysis of field-trial data for a collection of 25 WOSR genotypes demonstrated a significant impact of E on SY. We conclude that all traits, with the exception of SY, were influenced mostly by G and to a lesser, but statistically significant, extent by E as well as by GEI.

The F1_239 hybrid (G25) showed significant adaptability and stability with respect to BF and WTS, while hybrid F1_952 (G19) showed adaptability and stability for NSP and WTS. The WOSR cultivars Monolit (G01), Starter (G03), Mendel (G04) and Sherlock (G22), as well as the breeding line HO_TP (G17) and the Ogura hybrid F1_239 (G25), had the best TGSI. Two breeding lines, HO_SS and S43, were the least stable, having a low TGSI.