Received: 26 May 2023 Accepted: 14 September 2023 Published online: 2 October 2023 DOI: 10.1002/csc2.21106 Crop Science O R I G I N A L A R T I C L E C r o p B r e e d i n g & G e n e t i c s Exploring the trait-yield association patterns in different oat mega-environments of Canada Weikai Yan1 Mehri Hadinezhad1 Brad Dehaan1 Matt Hayes1 Savka Orozovic1 Kirby T. Nilsen2 Dan MacEachern3 Genevieve Telmosse4 Aaron Beattie5 Helen Booker6 Holly Byker7 Allan Cummiskey3 Isabelle Morasse4 Nathan Mountain8 Melinda Drummond8 Zhanghan Zhang6 Michael Holzworth9 Julie Durand10 Yuanhong Chen1 1Ottawa Research and Development Centre, AAFC, Ottawa, Ontario, Canada 2Brandon Research and Development Centre, AAFC, Brandon, Manitoba, Canada 3Charlottetown Research and Development Centre, AAFC, Charlottetown, Prince Edward Island, Canada 4Quebec Research and Development Center, AAFC, Normandin, Quebec, Canada 5Department of Plant Sciences, Crop Development Centre, University of Saskatchewan, Saskatoon, Saskatchewan, Canada 6Department of Plant Agriculture, Ontario Agricultural College (OAC), University of Guelph, Guelph, Ontario, Canada 7Northern and Eastern Ontario Crop Research Centre, University of Guelph, Winchester, Ontario, Canada 8New Liskeard Agricultural Research Station, University of Guelph, New Liskeard, Ontario, Canada 9University of Guelph Ridgetown Campus, Ridgetown, Ontario, Canada 10SemiCan Inc., Princeville, Quebec, Canada Correspondence Weikai Yan and Mehri Hadinezhad, Ottawa Research and Development Centre, AAFC, Ottawa, ON, Canada. Email: Weikai.yan@agr.gc.ca and Meri.hadinezhad@agr.gc.ca Assigned to Associate Editor Paulo Teodoro. Funding information Agriculture and Agri-Food Canada; Canadian Field Crops Research Alliance, Grant/Award Number: J-002090 Abstract This article presents a graphical method to visually analyze the trait–yield associa- tion (TYA) patterns based on data from multi-location, multi-year crop variety trials, exemplified using oat data from trials conducted across Canada from 2017 to 2022. Each year a new set of 60–66 oat (Avena sativa L.) breeding lines were tested in replicated yield trials at 9–11 locations, and data for yield, key agronomic and qual- ity traits, and crown rust scores were collected at all or some of the locations. Pearson correlation coefficient was calculated between yield and each trait for each trial. The correlation coefficients from different locations and years were arranged in a TYA × trial (TYT) two-way table. This table was subjected to singular value decomposi- tion, and the resulting first two principal components were used to generate a TYT biplot. The TYT biplot revealed three oat mega-environments (MEs) in Canada, con- sistent with the conclusion from previous ME analyses, indicating that each ME had its characteristic TYA patterns. It was found that yield was consistently and positively Abbreviations: AEC, average-environment-coordination; ME, mega-environment; ORDC, Ottawa Research and Development Centre of Agriculture and Agri-Food Canada; TYA, trait–yield associations; TYT, TYA × trial (two-way table or biplot). This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2023 His Majesty the King in Right of Canada and The Authors. Crop Science published by Wiley Periodicals LLC on behalf of Crop Science Society of America. Reproduced with the permission of the Minister of Agriculture and Agri-Food Canada. 3356 wileyonlinelibrary.com/journal/csc2 Crop Science. 2023;63:3356–3366. https://orcid.org/0000-0002-5704-1570 mailto:Weikai.yan@agr.gc.ca mailto:Meri.hadinezhad@agr.gc.ca http://creativecommons.org/licenses/by/4.0/ https://wileyonlinelibrary.com/journal/csc2 http://crossmark.crossref.org/dialog/?doi=10.1002%2Fcsc2.21106&domain=pdf&date_stamp=2023-10-02 YAN ET AL. 3357Crop Science correlated with crown rust (Puccinia coronata var. avenae) resistance, test weight, kernel weight, and groat content in ME1 (the crown rust-prone regions of Ontario); yield was correlated positively with plant height but negatively with oil content in ME2 (the northern regions of eastern Canada). Interestingly, crown rust resistance was found to contribute negatively to yield in ME2. No strong and consistent TYAs were found in ME3 (the Canadian prairies). Different types of TYAs have different uses in genotypic selection. 1 INTRODUCTION The breeding cycle of a crop such as oat (Avena sativa L.) consists of four major stages: parent selection and hybridiza- tion, generation advance, visual selection, and yield trials. In the Ottawa Research and Development Centre of Agriculture and Agri-Food Canada (ORDC) oat breeding program, about 98% of the original breeding population is eliminated through visual selection, and only 2% of the population have the oppor- tunity to be evaluated in yield trials (Yan, 2021). Therefore, improving the selection accuracy at the visual selection stage is critical to improve breeding efficiency. The breeding objectives for any crop and region can be clas- sified into two types of traits: yield as one and agronomic traits, quality, and disease/pest resistance as the other, and the ultimate goal is to combine high yield and desirable levels of various traits. Yield is always the most important breed- ing objective; the economic value of other traits has to build on the basis of the yield level (Yan & Frégeau-Reid, 2018). Therefore, understanding the nature and magnitude of the associations between yield and these other traits, referred to as trait–yield associations (TYA) hereafter, is crucial to select high-yielding, superior-quality, resilient cultivars both at the yield trial stage and particularly at the visual selection stage. At the yield trial stage, yield and other key traits can be directly measured and selected. The main challenge is to combine high yield with desirable levels of these traits (e.g., Howarth et al., 2021; Yan & Frégeau-Reid, 2018). When unfa- vorable association exists between yield and a key trait, which is often the case, compromise must be made to maximize an integrative selection index, such as a yield × trait index (Yan & Frégeau-Reid, 2018) or the more traditional linear indices (Baker, 2020; Céron-Rojas & Crossa, 2018; Olivoto et al., 2021; Yan, 2014). At the visual selection stage, in which direct selection for yield is not possible, selection for yield is through indirect selection for traits that are readily observable and thought to contribute to yield. This is an essential task with great uncer- tainties (Bowman et al., 2004; Dahiya et al., 1984; Ud-Din et al., 1993), and a good understanding of the various TYAs is crucial. Traits favorably associated with yield can be used in indirect selection for yield in addition to selection for the traits per se; traits unfavorably associated with yield should be selected with caution to prevent loss of high-yielding geno- types; traits not closely associated with yield can be used in independent culling with confidence. A good understanding of various TYAs can only be achieved through analysis of trait–yield data covering mul- tiple locations and years to take into account the ever-present genotype-by-environment interactions (Basford & Cooper, 1998; Ceccarelli et al., 1994; Cooper & Hammer, 1996; de Leon et al., 2016; Kang & Gauch, 1996). Variety trials are conducted every year for all major crops and regions. In such trials, yield and key agronomic, disease, and quality traits are typically measured. Such data provide an invaluable opportu- nity to understand the TYA patterns for the crop and region of interest. This article introduces a TYA × trial (TYT) biplot and demonstrates its use in visual analysis of TYA patterns across trials (location–year combinations) to reveal mega- environments (MEs); to visualize the nature, strength, and consistency of a TYA across locations and years; and to discuss the implications of various types of TYAs on the strategies of visual selection. 2 MATERIALS AND METHODS The ORDC oat preliminary yield test is conducted yearly at 9–11 locations across Canada. The data analyzed here were from the 2017 and 2019–2022 trials. Data from 2018 were excluded because data for crown rust (Puccinia coronata var. avenae) score, an important trait for the current study, were not available. The locations and trials involved in this study can be found in Table 1. Sixty to 66 new breeding lines plus several check cultivars were tested each year using an incomplete blocks design with two replications. The genotypes tested within a year were the same at all locations, but they were completely different in different years, except for a single-check cultivar. At most locations, the plots (the experimental units) were of four or six rows that were 3.5-m long. At La Poctiere and Princeville, Quebec, the plots were composed of eight rows of 5.0-m long. Locally recommended management practices were adopted at each location, except that no fungicide was used. Grain yield was determined at all locations. Agronomic traits, including days to heading, days 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 3358 YAN ET AL.Crop Science to maturity, plant height, and lodging scores, and grain qual- ity traits, including thousand kernel weight and test weight, were determined at most locations. Groat content, β-glucan content, oil content, and protein content were determined for 4–5 locations each year. Crown rust scores were recorded during grain filling wherever the disease occurred. The quality traits were determined using samples bulked across replications in a trial. The yield, agronomic traits, and crown rust scores were determined for each plot. Genotypic means for each trait in each trial were calculated by considering the experimental design and the within-block field trend using a polynomial smoothing method (Yan, 2014). Pearson correlation coefficient was calculated between the genotypic means of yield and each trait in each trial. Each trait–yield correlation is referred to as a TYA. The correla- tion table from all trials was then assembled into a TYA by trial two-way table in the form of Table 1. Since not all traits were determined in all trials, some TYAs were missing in some trials, which were filled using estimated values using a missing value imputation method (Yan, 2013). Since crown rust did not occur in most northern locations, the yearly mean crown rust scores for each genotype were calculated across locations where it did occur. These mean scores were then used to calculate the crown rust score by yield correlation in each trial (“CRUST_MEAN,”; Table 1). Preliminary analysis showed that genotypic ranking in crown rust scores within a year varied depending on the trials, reflecting different strain compositions of the pathogen at different locations. Never- theless, the mean crown rust scores across locations were considered a good representation of the genotypes’ crown rust susceptibility/resistance. The TYA by trial (TYT) two-way table of correlations was subjected to singular value decomposition, without any cen- tering or scaling ("Centering= 0, Scaling= 0,"; Figure 1), and the resulting first two principal components (PC1 and PC2) were used to construct a biplot (Gabriel, 1971), referred to as a TYT biplot. The analyses were conducted using GGEbiplot (Yan, 2014). 3 RESULTS AND DISCUSSION 3.1 How to interpret a TYT biplot The Pearson correlation coefficients between yield and each trait from each of the trials (location–year combinations) are presented in Table 1, which is approximately displayed in the TYT biplot in Figure 1. In Figure 1, each TYA is presented by the name or abbrevia- tion of the trait in question in blue, while each trial is presented as a location–year combination in red. Each TYA or trial is connected to the biplot origin. Concentric circles are drawn to facilitate visualization of the distance of each TYA or trial to Core Ideas ∙ A trait–yield association × trial (TYT) biplot was introduced to reveal trait–yield association patterns. ∙ TYT biplots revealed different oat mega- environments in Canada, each with unique trait–yield association patterns. ∙ Trait–yield association patterns can be used to guide visual selection. ∙ Crown rust resistance was shown to be counterpro- ductive in the non-rust regions. ∙ Different types of trait–yield associations have different uses in genotypic selection. the biplot origin, referred to as the vector of the TYA or trial. Although there are numerous ways to present a biplot (Yan & Kang, 2002; Yan & Tinker, 2006), as seen in numerous pub- lications, Figure 1 shows the very basic form. All other forms are but derivatives from this one. From Figure 1, the following pieces of information can be visualized: (1) The vector length of a TYA indicates the strength of the association between yield and the trait in ques- tion. For example, CRUST (correlation of crown rust scores with yield) had the longest vector, followed by KG/HL (cor- relation of test weight in kg/hL with yield), DTM (correlation of days to maturity with yield), HEIGHT (correlation of plant height with yield), and TKW (correlation of thousand kernel weight with yield), indicating that these traits expressed the strongest correlations with yield in at least some of the tri- als. On the contrary, TYAs with short vectors, for example, BGL (correlation of β-glucan with yield), indicate that the traits in question were not strongly associated with yield in any of the trials. (2) The vector length of a trial (i.e., environ- ment) indicates the strength of TYAs in the trial. For example, OTT_22 and PALM_19 showed long vectors, indicating that strong TYAs occurred in these trials. (3) The cosine of the angle between two TYAs indicates their similarity across the trials. For example, CRUST and CRUST_MEAN had a very small angle, indicating that they were highly similar across the trials. CRUST was the correlation between crown rust score and yield. Since crown rust did not occur in many of the tri- als, the correlation in these trials could not be calculated and had to be imputed. CRUST_MEAN was the calculated corre- lation between yield and the mean crown rust score. The high similarity between CRUST and CRUST_MEAN is expected; it also validates the missing value imputation method. (4) The cosine of the angle between two trials indicates the sim- ilarity of the trials in TYA patterns. For example, Figure 1 shows that OTT_22, PALM_19, and ELORA_22 were highly 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense YAN ET AL. 3359Crop Science T A B L E 1 Pearson correlations between various traits and yield in individual trials (location–year combinations). Trials DTM DTH Height Lodging score KG/HL TKW Groat content β-Glucan content Oil content Protein content Crown rust score Mean crown rust score Mega-environment 1 ELORA_21 0.36 −0.28 −0.07 −0.28 0.50 0.41 0.40 −0.18 −0.01 0.20 −0.58 −0.58 ELORA_22 0.65 −0.34 −0.13 −0.27 0.64 0.50 0.53 −0.25 0.02 0.26 −0.78 −0.78 OTT_17 0.05 0.10 −0.02 0.15 −0.15 −0.14 −0.10 0.03 0.02 −0.06 0.13 0.13 OTT_19 −0.04 −0.15 0.03 −0.21 0.22 0.21 0.16 −0.06 −0.02 0.09 −0.21 −0.21 OTT_20 −0.96 −0.24 0.28 −0.58 0.21 0.30 0.04 0.06 −0.13 0.06 0.05 0.04 OTT_21 0.24 −0.01 −0.06 0.05 0.06 0.02 0.07 −0.05 0.02 0.03 −0.12 −0.11 OTT_22 0.62 −0.34 −0.13 −0.28 0.64 0.50 0.53 −0.25 0.01 0.26 −0.78 −0.78 PALM_17 0.02 −0.14 0.01 −0.18 0.22 0.20 0.16 −0.06 −0.02 0.09 −0.22 −0.22 PALM_19 0.08 −0.49 0.03 −0.63 0.78 0.69 0.57 −0.22 −0.06 0.30 −0.77 −0.78 PALM_20 0.28 0.02 −0.08 0.11 0.01 −0.02 0.04 −0.04 0.03 0.01 −0.08 −0.08 PALM_22 0.20 −0.29 −0.02 −0.33 0.49 0.42 0.37 −0.16 −0.02 0.19 −0.53 −0.53 Mean 0.14 −0.20 −0.02 −0.22 0.33 0.28 0.25 −0.10 −0.01 0.13 −0.35 −0.36 Mega-environment 2 LAPO_17 0.20 0.24 0.14 0.07 −0.20 −0.11 −0.15 −0.20 −0.18 −0.15 0.06 0.03 LAPO_21 0.51 0.13 −0.24 0.15 −0.28 −0.48 −0.42 −0.10 0.24 0.03 −0.54 −0.29 NL_17 −0.13 0.03 0.13 −0.05 0.08 0.15 0.08 −0.02 −0.13 −0.03 0.12 0.10 NL_19 0.22 0.25 0.13 0.08 −0.23 −0.12 −0.14 −0.20 −0.18 −0.16 0.10 0.03 NL_20 0.00 0.32 0.38 0.03 −0.20 0.14 0.06 −0.27 −0.46 −0.30 0.56 0.26 NL_21 −0.32 0.10 0.37 −0.14 0.24 0.38 0.16 −0.09 −0.35 −0.08 0.22 0.24 NORM_17 0.11 0.35 0.33 0.07 −0.26 0.03 −0.02 −0.29 −0.42 −0.30 0.46 0.20 NORM_19 0.14 0.24 0.19 0.02 −0.10 −0.04 −0.15 −0.20 −0.20 −0.12 −0.03 0.05 NORM_21 −0.31 0.19 0.47 −0.11 0.12 0.40 0.21 −0.16 −0.48 −0.19 0.49 0.33 NORM_22 0.17 0.27 0.20 0.02 −0.10 −0.06 −0.20 −0.22 −0.21 −0.12 −0.10 0.03 PE_17 0.14 0.34 0.31 0.06 −0.22 0.01 −0.08 −0.29 −0.38 −0.26 0.30 0.16 PE_19 0.14 0.26 0.21 0.02 −0.10 −0.04 −0.16 −0.22 −0.22 −0.13 −0.03 0.05 PE_20 −0.30 0.05 0.31 −0.18 0.35 0.34 0.06 −0.05 −0.23 0.03 −0.12 0.15 PE_21 −0.21 0.14 0.33 −0.05 0.02 0.28 0.19 −0.12 −0.37 −0.18 0.50 0.26 PE_22 0.11 0.10 0.03 0.05 −0.14 −0.07 −0.04 −0.08 −0.07 −0.09 0.11 0.02 PRIN_19 0.20 0.24 0.14 0.06 −0.17 −0.11 −0.17 −0.20 −0.17 −0.13 −0.01 0.02 PRIN_20 −0.12 0.39 0.60 −0.20 0.37 0.29 −0.27 −0.35 −0.47 −0.06 −0.53 0.15 PRIN_22 0.11 0.05 −0.04 0.10 −0.26 −0.10 0.06 −0.03 −0.04 −0.13 0.38 0.04 Mean 0.04 0.20 0.22 0.00 −0.06 0.05 −0.06 −0.17 −0.24 −0.13 0.11 0.10 Mega-environment 3 BRA_17 −0.06 0.00 −0.09 0.05 −0.33 −0.10 −0.06 0.03 0.04 −0.16 0.20 BRA_19 0.86 0.87 −0.03 0.61 0.65 0.15 −0.56 0.14 −0.02 −0.52 0.12 BRA_20 0.30 0.29 −0.12 0.19 −0.42 −0.36 −0.03 0.02 0.26 −0.52 0.08 BRA_21 0.38 0.26 0.04 0.07 0.13 −0.22 0.13 −0.06 0.21 −0.22 −0.31 BRA_22 0.37 0.19 0.11 −0.02 0.32 −0.18 0.24 −0.10 0.20 −0.07 −0.49 LAC_17 0.10 0.17 −0.09 0.18 −0.14 −0.01 −0.21 0.07 0.00 −0.23 0.25 LAC_19 −0.25 −0.20 −0.03 −0.10 −0.22 0.01 0.03 0.00 −0.04 0.09 0.12 LAC_20 −0.15 −0.13 0.00 −0.08 −0.07 0.03 0.04 −0.01 −0.03 0.10 0.02 LAC_21 0.48 0.38 0.00 0.18 0.08 −0.23 0.01 −0.02 0.22 −0.37 −0.20 LAC_22 −0.03 −0.15 0.11 −0.20 0.19 −0.05 0.29 −0.10 0.06 0.21 −0.37 SASK_17 0.86 0.87 −0.03 0.61 0.65 0.15 −0.56 0.14 −0.02 −0.52 0.12 (Continues) 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 3360 YAN ET AL.Crop Science T A B L E 1 (Continued) Mega-environment 3 SASK_19 0.07 0.05 −0.04 0.01 −0.25 −0.20 0.07 −0.02 0.14 −0.19 −0.02 SASK_20 −0.01 0.01 −0.03 0.01 −0.11 −0.05 0.00 0.00 0.03 −0.06 0.04 SASK_21 0.20 0.18 −0.04 0.10 −0.14 −0.19 0.00 0.00 0.14 −0.26 −0.02 SASK_22 −0.07 −0.02 −0.05 0.03 −0.17 −0.01 −0.07 0.03 −0.02 −0.06 0.15 Mean 0.20 0.18 −0.02 0.11 0.01 −0.08 −0.05 0.01 0.08 −0.19 −0.02 Note: Some values were unavailable and were imputed using a missing value imputation method. Each trial is presented as the location code and the last two digits of the year, joint with “_”. For example, the trial at Ottawa in 2017 is presented as “OTT_17”. The threshold of correlation coefficient for p = 0.05 (n = 66) is 0.226. Abbreviations: DTH, days to heading; DTM, days to maturity; KG/HL, test weight in kg/hL; TKW, thousand kernel weight. similar, all being trials in the crown rust-prone regions of Ontario. It also showed that PE_20, PRIN_20, NORM_21, and NL_21 were highly similar, all being trials in the non-rust regions of eastern Canada. (5) The product of the vector length of a TYA, the vector length of a trial, and the cosine of the angle between the TYA and the trial approximates the Pear- son correlation of the TYA in the trial (Table 1). For example, Figure 1 suggests that KG/HL were highly and positively cor- related with yield in OTT_22, PALM_19, and ELORA_22, while CRUST were highly and negatively correlated with yield in these trials. Finally, the goodness of fit of the biplot was 62.3%, indicating that the patterns shown in the biplot are 62.3% true to the table it approximates. The patterns regard- ing the TYAs or trials with longer vectors are more accurately displayed in the biplot. 3.2 Trait–yield associations and mega-environments Figure 1 shows that the TYAs are placed in all directions, indi- cating that the traits varied greatly in their association with yield; some traits were similar (with acute angles between them) and some were contrasting (with obtuse angles between them) in their associations with yield. The trials are also placed in all directions, indicating that TYAs varied greatly depending on the trials and that no TYA was the same in all trials, reflecting TYA by environment interactions. The trials did show a trend to group by locations, how- ever. Using the method of Yan (2015, 2019), Figure 1 is transformed into Figure 2, in which the trials at a location in different years are displayed as a cluster, the center being the mean coordination of all trials at the location and the mem- bers being the individual trials indicated by the last two digits of the years. From Figure 2, the locations appear to fall into three groups, corresponding to the three oat MEs in Canada (Yan et al., 2021), namely, OTT, ELORA, and PALM form one group, representing ME1 (the crown rust-prone regions of Ontario); NL, PE, PRIN, and NORM form the second group, represent- ing ME2 (the northern regions of eastern Canada); and BRA, LAC, and SASK form the third group, representing ME3 (the Canadian prairies). An exception is LAPO, which is a loca- tion in Quebec but fell into ME3. This location was repeatedly shown to differ from other geographically similar locations in cultivar responses (Yan, 2015, 2021; Yan et al., 2011). The ME differentiation in Figure 2 indicates that different MEs expressed different TYA patterns. 3.3 Trait–yield association patterns in ME1 In Figure 3, the TYT biplot for ME1 is presented. The most prominent patterns in this biplot include (1) OTT_20 was different from other trials; this was probably because the trial was seeded 2 weeks later than in normal years due to the COVID-19 pandemic. (2) CRUST or CRUST-MEAN was among the strongest TYAs. It had an obtuse angle with all trials except OTT_20, indicating a negative correlation between crown rust scores and yield in most of the tri- als. This result is consistent with the common knowledge that crown rust is a key yield determinant and crown rust resistance has been a key breeding objective for this ME (e.g., Yan et al., 2022a). (3) Opposite to CRUST, KG/HL was positive in most trials, and GROAT (groat content- yield association) and TKW (kernel weight-yield association) were similar to KG/HL but to a lesser extent. These TYAs suggest that the yield in ME1 can be improved by improv- ing crown rust resistance, test weight, kernel weight, and groat content. More importantly, these traits can be effec- tively observed and selected in the visual selection stage at Ottawa, a ME1 location. Figure 3 also shows that DTM (days to maturity) was negatively associated with yield in OTT_20 because the trial was seeded late, and late matu- rity became a key yield-limiting factor. Further, Figure 3 shows that plant height (HEIGHT), β-glucan content (BGL), oil content (OIL), and protein content (PROTEIN) were not strongly associated with yield in any of the trials. This sug- gests that desired levels of these traits can be relatively easily combined with high yield in ME1 and that these traits can be effectively used in independent culling for ME1. 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense YAN ET AL. 3361Crop Science The average-environment-coordination (AEC) form of Figure 3 is presented in Figure 4. The arrow on the AEC abscissa (the single arrowed line) points to mean correlation between yield and the trait in question and the AEC ordinate (the double arrowed line) separates traits with positive mean correlations with yield from those with negative mean corre- lations with yield. It shows that on average crown rust score (CRUST) was most negatively correlated with yield, followed by lodging (LOD) and days to heading (DTH), while test weight (KG/HL) was most positively associated with yield, followed by GROAT and TKW. It also shows that DTM had strong negative correlation with yield in OTT_20 but not in other trials. The patterns shown in the TYT biplot can be used as a guide to more closely examine the correlation coefficients in the TYT table (Table 1). 3.4 Trait–yield association patterns in ME2 The TYT biplot for ME2 is presented in Figure 5. The fol- lowing patterns can be observed. (1) The trial LAPO_21 was distinct from other trials in TYA patterns. This is consistent with the previous finding that LAPO behaved differently from other ME2 locations even though they were geographically F I G U R E 1 Trait–yield association × trial (TYT) biplot to show the trait–yield correlations in multi-location multi-year trials. Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hl. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: ELORA, Elora ON; OTT, Ottawa ON; PALM, Palmerston ON; LAPO, La Poctiere QC; NL, New Liskeard ON; NORM, Normandin QC; PE, Harrington PE; PRIN, Princeville QC; BRA, Brandon MB; LAC, Lacombe AB; SASK, Saskatoon SK. BGL, β-glucan content; GROAT, groat content; OIL, oil content. F I G U R E 2 Trait–yield association × trial (TYT) biplot to show mega-environments based on trait–yield associations. Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: ELORA, Elora ON; OTT, Ottawa ON; PALM, Palmerston ON; LAPO, La Poctiere QC; NL, New Liskeard ON; NORM, Normandin QC; PE, Harrington PE; PRIN, Princeville QC; BRA, Brandon MB; LAC, Lacombe AB; SASK, Saskatoon SK. BGL, β-glucan content; GROAT, groat content; OIL, oil content. close. (2) Plant height (HEIGHT) was positively associated with yield in all trials except LAPO_21. Interestingly, oat plants tended to be shorter in ME2 than in ME1, leading to less lodging in ME2 compared to ME1 (Table 2). This suggests that selection for shorter genotypes, as a means to improve lodging resistance, should be relaxed when ME2 is targeted. Taller genotypes that meet the criteria for other traits should be given an opportunity to show their yielding ability and lodging resistance in yield trials. A documented exam- ple is the cultivar "AAC Excellence" (Yan et al., 2022b), which is relatively tall but has demonstrated high yield, high β-glucan content, and superior grain quality in ME2. (3) Oil content was negatively associated with yield in all trials except LAPO_21, consistent with the correlated change of oil and grain yield in a recurrent selection study (Frey & Holland, 1999). This is a favorable TYA as low oil content is a pre- ferred trait for the milling industry. It suggests that selection for low oil content, a trait that is highly heritable (Baker & McKenzie, 1972; Frey & Holland, 1999; Yan et al., 2016), can result in high-yielding genotypes in ME2. (4) Contrasting to ME1, crown rust score was positively correlated or unas- sociated with yield in most trials in ME2. This suggests that crown rust resistance genes tended to contribute negatively to 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 3362 YAN ET AL.Crop Science T A B L E 2 Mean trait values across genotypes and years for each location and mega-environment. Location Yield (kg ka−1) DTH (days) DTM (days) Height (cm) Lodging (0–9) Test weight (kg hL−1) TKW (g) Groat (%) β-Glucan (%) Oil (%) Protein (%) Crown rust (0–9) Mega-environment 1 ELORA 4120 64 93 6.9 32.8 24.9 5.5 OTT 3776 59 83 0.5 54.7 35.0 65.0 4.7 7.4 14.7 4.5 PALM 4237 59 119 2.5 44.9 34.9 66.1 4.8 7.6 14.6 2.4 Mean 4045 61 93 101 3.3 44.1 31.6 65.5 4.8 7.5 14.7 4.2 Mega-environment 2 NL 5440 53 91 91 3.9 43.2 32.2 70.1 4.7 7.4 13.5 2.2 NORM 7502 106 89 1.6 57.2 37.7 71.2 4.5 7.4 14.0 PE 4539 68 99 46.3 38.0 69.5 4.4 7.8 12.2 PRIN 4788 60 94 2.9 51.5 34.5 Mean 5567 61 97 93 2.8 49.6 35.6 70.3 4.5 7.5 13.2 2.2 Mega-environment 3 BRA 6528 58 91 103 2.1 53.1 37.3 71.0 4.9 7.1 16.2 LAC 7982 57 105 117 3.2 58.1 39.3 72.7 4.7 7.5 15.7 SASK 3954 55 84 86 53.7 33.8 70.8 4.4 7.3 15.8 Mean 6154 57 93 102 2.7 55.0 36.8 71.5 4.7 7.3 15.9 Abbreviations: DTH, days to heading; DTM, days to maturity; TKW, thousand kernel weight. yield in trials where crown rust was absent. The finding that CRUST was negative in ME1 and positive in ME2 was consis- tent with that by Holland and Munkvold (2001), who studied the correlation between crown rust resistance and grain yield across oat genotypes that differed in crown rust resistance under fungicide-treated and untreated conditions. This obser- vation appears to support the hypothesis that the expression of disease resistance genes is an energy-demanding process and their expression is at the expense of yield (Comeau et al., 2010). Isogenic lines that differ only in crown rust resis- tance genes, such as those described in Rines et al. (2018), are needed to critically test this hypothesis. Practically, how- ever, since crown rust is not a limiting factor in ME2, the selection for crown rust resistance should be relaxed when ME2 is targeted. (4) DTH was positively associated with yield in many of the ME2 trials, suggesting that visual selection for early heading should also be relaxed. Thus, to develop high-yielding cultivars for ME2, tall, late, or rust-susceptible genotypes should be retained in visual selection if they appear to have good straw strength, yield potential, and grain qual- ity. The association of lodging score and yield (LOD) had a very short vector, indicating that superior lodging resistance can be relatively easily combined with high yield and lodging can be vigorously used in independent culling when ME2 is targeted. The AEC form of the biplot (Figure 6) provided a simplified summary of the TYAs in ME2. It highlights the strong and positive correlations of DTH and plant height with yield and the strong and negative correlation between oil content and yield. 3.5 Trait–yield association patterns in ME3 In Figure 7 is the TYT biplot for ME3 is presented. The trials are placed in all directions, indicating that no single TYA was consistent across locations and years. Some relatively strong associations included the negative correlation between pro- tein content and yield (PROTEIN), the positive correlation between days to heading and yield (DTH), and the positive correlation between days to maturity and yield (DTM) in some trials, all being unfavorable. Therefore, these TYAs are not useful in visual selection for yield for ME3. Figure 7 shows that oil (OIL), β-glucan content (BGL), kernel weight (TKW), and plant height (HEIGHT) were not strongly associated with yield in any of the trials. This information is useful; it suggests that these traits can be vigorously used in independent culling in visual selection when ME3 is targeted and that desired lev- els of these traits can be relatively easily combined with high yield in ME3. 3.6 Strategies of selection for yield prior to yield trials In the ORDC oat breeding program, around 98% of the breeding population (of ca. 10,000 lines) is eliminated through 2 years of visual selection at a single location; only about 2% of the population will have an opportunity to be tested in yield trials (Yan, 2021). Visual selection includes selection in the field for plant height, lodging, maturity, any diseases that occur, and plant and panicle size, and selection 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense YAN ET AL. 3363Crop Science F I G U R E 3 Trait–yield association × by trial (TYT) biplot to show the trait–yield correlations in trial in mega-environment-1 (ME1). Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST: crown rust score; DTH: days to heading; DTM: days to maturity; LOD: lodging score; TKW: thousand kernel weight; KG/HL: test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: ELORA: Elora ON; OTT: Ottawa ON, PALM: Palmerston ON. BGL, β-glucan content; GROAT, groat content; OIL, oil content. F I G U R E 4 The average-environment-coordination (AEC) form of the trait–yield association × trial (TYT) biplot for mega-environment-1 (ME1). Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: ELORA, Elora ON; OTT, Ottawa ON; PALM, Palmerston ON. BGL, β-glucan content; GROAT, groat content; OIL, oil content. F I G U R E 5 Trait–yield association × trial (TYT) biplot to show the trait–yield correlations in trial in mega-environment-2 (ME2). Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: NL, New Liskeard ON; NORM, Normandin QC; PE, Harrington PE; PRIN, Princeville QC. BGL, β-glucan content; GROAT, groat content; LAPO, La Poctiere QC; OIL, oil content. F I G U R E 6 The average-environment-coordination (AEC) form of the trait–yield association × trial (TYT) biplot for mega-environment-2 (ME2). Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: NL, New Liskeard ON; NORM, Normandin QC; PE, Harrington PE; PRIN, Princeville QC. BGL, β-glucan content; GROAT, groat content; LAPO, La Poctiere QC; OIL, oil content. 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense 3364 YAN ET AL.Crop Science F I G U R E 7 Trait–yield association × trial (TYT) biplot to show the trait–yield correlations in trial in mega-environment-3 (ME3). Each trait–yield association (TYA) is presented by the name or abbreviation of the trait. The trait abbreviations are as follows: CRUST, crown rust score; DTH, days to heading; DTM, days to maturity; LOD, lodging score; TKW, thousand kernel weight; KG/HL, test weight in kg/hL. Each trial is presented as the location abbreviation and the last two digits of the year. The location codes are: BRA, Brandon MB; LAC, Lacombe AB; SASK, Saskatoon SK. BGL, β-glucan content; GROAT, groat content; OIL, oil content. after harvest through visual examination of seed samples for kernel weight, test weight, plumpness, uniformity of kernels, hull, color, presence of awns, etc. A good understanding of the TYAs through analyses such as those demonstrated in this article will help improve the accuracy of visual selection. Although yield is the most important breeding objective, it cannot be selected directly in the visual selection stage. It can be selected only through traits that are observable and have strong and consistent associations with yield, such as the TYAs identified for ME1 (Figure 3). On the other hand, a large proportion of the breeding population is eliminated through independent culling for traits that are not strongly associated with yield, including the TYAs with short vectors as shown in the biplots (Figures 1,3,5, and 7). Indirect selection for yield based on strong TYAs (e.g., crown rust resistance for ME1) and independent culling based on weak TYAs (e.g., kernel weight and plant height for ME3) have been effective in developing superior crop cultivars in the past (e.g., Yan et al., 2023). However, selection based on visible characters alone is not effective enough to keep pace with the increasing population and increased need for food. As models of genomic selection (GS) (Bekele et al., 2020; Campbell et al., 2021; Heffner et al., 2009; Jannink et al., 2010) become mature, the yield of thousands of individuals can be predicted for each ME by genetic markers with reasonable cost and considerable accuracy. A combined, rather than mutually exclusive, use of visual selection for observable traits and GS prediction for yield will improve the selection accuracy in the pre-yield trial stage and reduce the number of genotypes that must be tested in, and therefore, cost of, yield trials. 4 CONCLUSIONS Consistent with the results from ME analysis of yield data (Yan et al., 2021), the oat growing regions in Canada were grouped into three MEs based on trait–yield associations across locations and years. This demonstrates that different MEs had different trait–yield association patterns, which can be used to guide visual selection for each ME prior to yield tri- als. In ME1, the crown rust prone regions of Ontario, crown rust resistance, test weight, kernel weight, and groat content were strongly and consistently correlated with yield. There- fore, visual selection for these traits will improve yield. In ME2, the northern regions of eastern Canada, plant height and, to a lesser extent, DTH were positively correlated with yield, while crown rust resistance was negatively correlated with yield. Therefore, visual selection for early heading, shorter plants, and crown rust resistance should be relaxed to prevent high-yielding genotypes from being discarded. No strong and consistent trait–yield associations were found for ME3, the Canadian prairies. On the contrary, a number of traits, including plant height, kernel weight, and oil and β- glucan contents, were found to be independent of yield; these traits can be effectively used in independent culling without impacting yield. AU T H O R C O N T R I B U T I O N S Weikai Yan: Conceptualization; data curation; fund- ing acquisition; methodology; supervision; visualization; writing—original draft; writing—review and editing. Mehri Hadinezhad: Conceptualization; methodology; supervision; writing—review and editing. Brad dehaan, Matt Hayes, Savka Orozovic, Allan Cummiskey, Isabelle Morasse, Nathan Mountain, Melinda Drummond, Zhanghan Zhang, Michael Holzworth, Julie Durand, and Yuanhong Chen: Methodology; project administration; writing— review and editing. Kirby T. Nilsen, Aaron Beattie, and Helen Booker: Conceptualization; methodology; super- vision; writing—review and editing. Dan MacEachern, Genevieve Telmosse, and Holly Byker: Methodology; project administration; supervision; writing—review and editing. A C K N O W L E D G M E N T S This work was supported by Agriculture and Agri-Food Canada (AAFC) and the Canadian Field Crops Research Alliance (CFCRA). We thank Wes Dyke, Kali Stewart, Josée 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense YAN ET AL. 3365Crop Science Lévesque, Céline Georlette, Florence Vachon-Laberge, and Martin Trembley for conducting some of the trials. C O N F L I C T O F I N T E R E S T S T AT E M E N T The authors declare no conflicts of interest. O R C I D Mehri Hadinezhad https://orcid.org/0000-0002-5704- 1570 R E F E R E N C E S Baker, R. J. (2020). Selection indices in plant breeding. CRC Press. Baker, R. J., & McKenzie, R. I. H. (1972). Heritability of oil content in oats, Avena sativa L.1. Crop Science, 12(2), 201–202. https://doi.org/ 10.2135/cropsci1972.0011183X001200020015x Basford, K. E., & Cooper, M. (1998). Genotype× environment interac- tions and some considerations of their implications for wheat breeding in Australia. Australian Journal of Agricultural Research, 49(2), 153–174. https://doi.org/10.1071/A97035 Bekele, W. A., Itaya, A., Boyle, B., Yan, W., Mitchell Fetch, J., & Tinker, N. A. (2020). A targeted genotyping-by-sequencing tool (rapture) for genomics-assisted breeding in oat. Theoretical and Applied Genetics, 133, 653–664. https://doi.org/10.1007/s00122-019-03496-w Bowman, D. T., Bourland, F. M., Myers, G. O., Wallace, T. P., & Caldwell, D. (2004). Visual selection for yield in cotton breeding programs. The Journal of Cotton Science, 8(2), 62–68. Campbell, M. T., Hu, H., Yeats, T. H., Brzozowski, L. J., Caffe-Treml, M., Gutiérrez, L., Smith, K. P., Sorrells, M. E., Gore, M. A., & Jannink, J. L. (2021). Improving genomic prediction for seed quality traits in oat (Avena sativa L.) using trait-specific relationship matri- ces. Frontiers in Genetics, 12, 643733. https://doi.org/10.3389/fgene. 2021.643733 Ceccarelli, S., Erskine, W., Hamblin, J., & Grando, E. S. (1994). Genotype by environment interaction and international breeding pro- grammes. Experimental Agriculture, 30(2), 177–187. https://doi.org/ 10.1017/S0014479700024121 Céron-Rojas, J. J., & Crossa, J. (2018). Linear selection indices in modern plant breeding (pp. 256). Springer Nature. Comeau, A., Caetano, V. R., Haber, S., Langevin, F., Lévesque, M., & Gilbert, J. (2010). Systemic heuristic approaches guide the interaction of enhanced genetic diversity and complex stresses to generate bet- ter wheat germplasm faster and at lower cost. In I. Kovalchuk, & O. Kovalchuk, (Eds.), Genome instability and transgenerational effects (pp. 401–446). Nova Science Publisher. Cooper, M., & Hammer, G. L. (Eds.). (1996). Plant adaptation and crop improvement. IRRI. Dahiya, B. S., Waldia, R. S., Kaushik, L. S., & Solanki, I. S. (1984). Early generation yield testing versus visual selection in chickpea (Cicer arietinum L.). Theoretical and Applied Genetics, 68, 525–529. https://doi.org/10.1007/BF00285005 de Leon, N., Jannink, J. L., Edwards, J. W., & Kaeppler, S. M. (2016). Introduction to a special issue on genotype by environment interaction. Crop Science, 56(5), 2081–2089. https://doi.org/10.2135/ cropsci2016.07.0002in Frey, K. J., & Holland, J. B. (1999). Nine cycles of recurrent selection for increased groat-oil content in oat. Crop Science, 39(6), 1636–1641. https://doi.org/10.2135/cropsci1999.3961636x Gabriel, K. R. (1971). The biplot graphic display of matrices with appli- cation to principal component analysis. Biometrika, 58(3), 453–467. https://doi.org/10.1093/biomet/58.3.453 Heffner, E. L., Sorrells, M. E., & Jannink, J. L. (2009). Genomic selec- tion for crop improvement. Crop Science, 49(1), 1–12. https://doi.org/ 10.2135/cropsci2008.08.0512 Holland, J. B., & Munkvold, G. P. (2001). Genetic relationships of crown rust resistance, grain yield, test weight, and seed weight in oat. Crop Science, 41(4), 1041–1050. https://doi.org/10.2135/ cropsci2001.4141041x Howarth, C. J., Martinez-Martin, P. M. J., Cowan, A. A., Griffiths, I. M., Sanderson, R., Lister, S. J., Langdon, T., Clarke, S., Fradgley, N., & Marshall, A. H. (2021). Genotype and environment affect the grain quality and yield of winter oats (Avena sativa L.). Foods, 10, 2356. https://doi.org/10.3390/foods10102356 Jannink, J. L., Lorenz, A. J., & Iwata, H. (2010). Genomic selection in plant breeding: From theory to practice. Briefings in Functional Genomics, 9(2), 166–177. https://doi.org/10.1093/bfgp/elq001 Kang, M. S., & Gauch, H. G. (1996). Genotype-by-environment interac- tion. CRC press. Olivoto, T., Nardino, M., Meira, D., Meier, C., Follmann, D. N., de Souza, V. Q., Konflanz, V. A., & Baretta, D. (2021). Multi-trait selec- tion for mean performance and stability in maize. Agronomy Journal, 113(5), 3968–3974. https://doi.org/10.1002/agj2.20741 Rines, H. W., Miller, M. E., Carson, M., Chao, S., Tiede, T., Wiersma, J., & Kianian, S. F. (2018). Identification, introgression, and molecular marker genetic analysis and selection of a highly effective novel oat crown rust resistance from diploid oat, Avena strigosa. Theoretical and Applied Genetics, 131, 721–733. https://doi.org/10.1007/s00122- 017-3031-0 Ud-Din, N., Carver, B. F., & Krenzer, E. G., Jr. (1993). Visual selection for forage yield in winter wheat. Crop Science, 33(1), 41–45. https:// doi.org/10.2135/cropsci1993.0011183X003300010005x Yan, W. (2013). Biplot analysis of incomplete two-way data. Crop Science, 53(1), 48–57. https://doi.org/10.2135/cropsci2012.05.0301 Yan, W. (2014). Crop variety trials: Data management and analysis. John Wiley & Sons. Yan, W. (2015). Mega-environment analysis and test location evaluation based on unbalanced multiyear data. Crop Science, 55(1), 113–122. https://doi.org/10.2135/cropsci2014.03.0203 Yan, W. (2019). LG biplot: A graphical method for mega-environment investigation using existing crop variety trial data. Scientific Reports, 9(1), 7130. https://doi.org/10.1038/s41598-019-43683-9 Yan, W. (2021). A systematic narration of some key concepts and pro- cedures in plant breeding. Frontiers in Plant Science, 12, 724517. https://doi.org/10.3389/fpls.2021.724517 Yan, W., & Frégeau-Reid, J. (2018). Genotype by yield* trait (GYT) biplot: A novel approach for genotype selection based on multiple traits. Scientific Reports, 8(1), 8242. https://doi.org/10.1038/s41598- 018-26688-8 Yan, W., Frégeau-Reid, J., DeHaan, B., Thomas, S., Hayes, M., Martin, R., Cummiskey, A., Pageau, D., Morasse, I., Orozovic, S., Mitchell- Fetch, J., Menzies, J., Xue, A., & Mountain, N. (2022a). AAC reid oat. Canadian Journal of Plant Science, 102(3), 781–784. https://doi. org/10.1139/cjps-2022-0009 Yan, W., Frégeau-Reid, J., DeHaan, B., Thomas, S., Hayes, M., Martin, R., Cummiskey, A., Pageau, D., Morasse, I., Orozovic, S., Mitchell- Fetch, J., Menzies, J., Xue, A., & Mountain, N. (2022b). AAC 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense https://orcid.org/0000-0002-5704-1570 https://orcid.org/0000-0002-5704-1570 https://orcid.org/0000-0002-5704-1570 https://doi.org/10.2135/cropsci1972.0011183X001200020015x https://doi.org/10.2135/cropsci1972.0011183X001200020015x https://doi.org/10.1071/A97035 https://doi.org/10.1007/s00122-019-03496-w https://doi.org/10.3389/fgene.2021.643733 https://doi.org/10.3389/fgene.2021.643733 https://doi.org/10.1017/S0014479700024121 https://doi.org/10.1017/S0014479700024121 https://doi.org/10.1007/BF00285005 https://doi.org/10.2135/cropsci2016.07.0002in https://doi.org/10.2135/cropsci2016.07.0002in https://doi.org/10.2135/cropsci1999.3961636x https://doi.org/10.1093/biomet/58.3.453 https://doi.org/10.2135/cropsci2008.08.0512 https://doi.org/10.2135/cropsci2008.08.0512 https://doi.org/10.2135/cropsci2001.4141041x https://doi.org/10.2135/cropsci2001.4141041x https://doi.org/10.3390/foods10102356 https://doi.org/10.1093/bfgp/elq001 https://doi.org/10.1002/agj2.20741 https://doi.org/10.1007/s00122-017-3031-0 https://doi.org/10.1007/s00122-017-3031-0 https://doi.org/10.2135/cropsci1993.0011183X003300010005x https://doi.org/10.2135/cropsci1993.0011183X003300010005x https://doi.org/10.2135/cropsci2012.05.0301 https://doi.org/10.2135/cropsci2014.03.0203 https://doi.org/10.1038/s41598-019-43683-9 https://doi.org/10.3389/fpls.2021.724517 https://doi.org/10.1038/s41598-018-26688-8 https://doi.org/10.1038/s41598-018-26688-8 https://doi.org/10.1139/cjps-2022-0009 https://doi.org/10.1139/cjps-2022-0009 3366 YAN ET AL.Crop Science excellence oat. Canadian Journal of Plant Science, 102(3), 776–780. https://doi.org/10.1139/cjps-2021-0276 Yan, W., Frégeau-Reid, J., Pageau, D., & Martin, R. (2016). Genotype- by-environment interaction and trait associations in two genetic populations of oat. Crop Science, 56(3), 1136–1145. https://doi.org/ 10.2135/cropsci2015.11.0678 Yan, W., & Kang, M. S. (2002). GGE biplot analysis: A graphical tool for breeders, geneticists, and agronomists. CRC press. Yan, W., Mitchell-Fetch, J., Beattie, A., Nilsen, K. T., Pageau, D., DeHaan, B., Hayes, M., Mountain, N., Cummiskey, A., & MacEachern, D. (2021). Oat mega-environments in Canada. Crop Science, 61(2), 1141–1153. https://doi.org/10.1002/csc2.20426 Yan, W., Nilsen, K. T., & Beattie, A. (2023). Mega-environment analysis and breeding for specific adaptation. Crop Science, 63(2), 480–494. https://doi.org/10.1002/csc2.20895 Yan, W., Pageau, D., Frégeau-Reid, J., Lajeunesse, J., Goulet, J., Durand, J., & Marois, D. (2011). Oat mega-environments and test-locations in Quebec. Canadian Journal of Plant Science, 91(4), 643–649. https:// doi.org/10.4141/cjps10139 Yan, W., & Tinker, N. A. (2006). Biplot analysis of multi-environment trial data: Principles and applications. Canadian Journal of Plant Science, 86(3), 623–645. https://doi.org/10.4141/P05-169 How to cite this article: Yan, W., Hadinezhad, M., Dehaan, B., Hayes, M., Orozovic, S., Nilsen, K. T., MacEachern, D., Telmosse, G., Beattie, A., Booker, H., Byker, H., Cummiskey, A., Morasse, I., Mountain, N., Drummond, M., Zhang, Z., Holzworth, M., Durand, J., & Chen, Y. (2023). Exploring the trait-yield association patterns in different oat mega-environments of Canada. Crop Science, 63, 3356–3366. https://doi.org/10.1002/csc2.21106 14350653, 2023, 6, D ow nloaded from https://acsess.onlinelibrary.w iley.com /doi/10.1002/csc2.21106 by C anadian A griculture L ibrary, W iley O nline L ibrary on [22/04/2024]. See the T erm s and C onditions (https://onlinelibrary.w iley.com /term s-and-conditions) on W iley O nline L ibrary for rules of use; O A articles are governed by the applicable C reative C om m ons L icense https://doi.org/10.1139/cjps-2021-0276 https://doi.org/10.2135/cropsci2015.11.0678 https://doi.org/10.2135/cropsci2015.11.0678 https://doi.org/10.1002/csc2.20426 https://doi.org/10.1002/csc2.20895 https://doi.org/10.4141/cjps10139 https://doi.org/10.4141/cjps10139 https://doi.org/10.4141/P05-169 https://doi.org/10.1002/csc2.21106 Exploring the trait-yield association patterns in different oat mega-environments of Canada Abstract 1 | INTRODUCTION 2 | MATERIALS AND METHODS 3 | RESULTS AND DISCUSSION 3.1 | How to interpret a TYT biplot 3.2 | Trait-yield associations and mega-environments 3.3 | Trait-yield association patterns in ME1 3.4 | Trait-yield association patterns in ME2 3.5 | Trait-yield association patterns in ME3 3.6 | Strategies of selection for yield prior to yield trials 4 | CONCLUSIONS AUTHOR CONTRIBUTIONS ACKNOWLEDGMENTS CONFLICT OF INTEREST STATEMENT ORCID REFERENCES