INTEGRATING FACTOR ANALYTIC MODELS AND ENVIROTYPING FOR PREDICTING SOYBEAN PERFORMANCE IN MULTI-ENVIRONMENT TRIALS

Maurício dos Santos Araújo; Saulo Fabrício da Silva Chaves; Luiz Antônio dos Santos Dias; André R. G. Bezerra; Germano Costa Neto; Kaio Olimpio das Graças Dias

Todos os Trabalhos

Trabalho

Título do Trabalho

INTEGRATING FACTOR ANALYTIC MODELS AND ENVIROTYPING FOR PREDICTING SOYBEAN PERFORMANCE IN MULTI-ENVIRONMENT TRIALS

Autores

Maurício dos Santos Araújo
Saulo Fabrício da Silva Chaves
Luiz Antônio dos Santos Dias
André R. G. Bezerra
Germano Costa Neto
Kaio Olimpio das Graças Dias

Modalidade

Prêmio Estudantes Alcides de Carvalho (nível doutorado)

Área temática

Biometria Aplicada ao Melhoramento de Plantas

Data de Publicação

21/09/2023

País da Publicação

Brasil

Idioma da Publicação

Português

Página do Trabalho

https://www.even3.com.br/anais/12cbmp/654559-integrating-factor-analytic-models-and-envirotyping-for-predicting-soybean-performance-in-multi-environment-trial

ISBN

978-85-5722-993-8

Palavras-Chave

Enviromics, envirotype, and mixed models.

Resumo

Soybean, a critical crop globally, presents variable productivity across different cultivation regions due to genotype-by-environment interaction (GEI). Environmental covariates (EC's) allows the characterization of target population environments (TPE), identifying areas where genotypes can maximize yield. The aimed was to select top 15% of genotypes for high performance and stability using analytical factor models (FA). Moreover, we aimed to predict genotypes using ECs data, subsequently draw thematic maps to recommend cultivars in untested environments. One hundred and ninety-five genotypes were evaluated using a randomized complete block design with three repetitions over three years (2019-2021), across 49 environments in the state of Mato Grosso do Sul. Thirteen statistical models were analyzed to ascertain the existence of spatial trends, using the first-order autoregressive model. We selected the best-fitted model for each environment, which we then applied to a single-step multi-environment analysis via FA model. Genotype selection was conducted via factor analytic selection tools (FAST), using the parameters of overall performance (OP) and Root Mean Square Deviation (RMSD). We calculated a selection index considering high OP and lower RMSD, along with individual accuracy per genotype. The best-fitted FA model was chosen based on AIC, explained total variance, and average semivariances ratio. Thirty-two ECs were sourced from the NASA Power and SoilGrids databases. We employed Euclidean distance plus ECs to characterize both the conducted and to-be-predicted environments. We applied the interpolation method using inverse distance weighting. The prediction for non-evaluated environments was performed using the kernel of Partial least squares regression. Leave-one-out cross validation was using for dataset, and genotype recommendations based on thematic maps: (i) specific adaptation map, (ii) which-won-check; (iii) which-won-where; and (iv) pair-to-pair comparison. FA4 model explained 84.76 and 76.63% of total variance and covariance, respectively. The environments evaluated outside the TPE showed high similarities with the experimental trials, indicating high reliability. GEI partitioning showed 81% is of complex type, and the majority of genetic correlations between environments were positive. G177, G100 G144, G088, and G16 had high overall performance, with the latter two having high reliability. G178, G031, G101, G052, G035, G144 were stable (lower RMSD). Prediction accuracy was 0.58, 0.76, and 0.77 for eBLUE, eBLUP, and loading factor, respectively. Map (i) showed that G088 achieved high OP (322.9) with a predicted yield range varying from 3751-4000 and >4000 (kg ha-1) throughout the state of MS. Map (ii) showed that G016 surpassed check C054 in most pixels of the state. In map (iii), when comparing G100 and G177, both genotypes demonstrated specific adaptations for the state. Map (iv) suggested the formation of four mega-environments. This approach holds promise for plant breeding, as using ECs can predict the behavior of genotypes in non-evaluated environments, mapping regions that increase each genotype's productive potential.

Título do Evento: 12º Congresso Brasileiro de Melhoramento de Plantas 2023
Cidade do Evento: Caxambu
Título dos Anais do Evento: Anais do 12º Congresso Brasileiro de Melhoramento de Plantas
Nome da Editora: Even3
Meio de Divulgação: Meio Digital

Como citar

ARAÚJO, Maurício dos Santos et al.. INTEGRATING FACTOR ANALYTIC MODELS AND ENVIROTYPING FOR PREDICTING SOYBEAN PERFORMANCE IN MULTI-ENVIRONMENT TRIALS.. In: Anais do 12º Congresso Brasileiro de Melhoramento de Plantas. Anais...Caxambu(MG) Hotel Glória, 2023. Disponível em: https//www.even3.com.br/anais/12cbmp/654559-INTEGRATING-FACTOR-ANALYTIC-MODELS-AND-ENVIROTYPING-FOR-PREDICTING-SOYBEAN-PERFORMANCE-IN-MULTI-ENVIRONMENT-TRIAL. Acesso em: 07/06/2025