ABSTRACT Traditional germination tests which assess seed quality are costly and time-consuming, mainly when performed on a large scale. In this study, we assessed the efficiency of X-ray imaging analyses in predicting the physiological quality of tomato seeds. A convolutional neural network (CNN) called mask region convolutional neural network (MaskRCNN) was also tested for its precision in adequately classifying tomato seeds into four seed quality categories. For this purpose, X-ray images were taken of seeds of 49 tomato genotypes (46 Solanum pennellii introgression lines) from two different growing seasons. Four replicates of 25 seeds for each genotype were analyzed. These seeds were further assessed for germination and seedling vigor-related traits in two independent trials. Correlation analysis revealed significant linear association between germination and image-based variables. Most genotypes differed in terms of germination and seed development performance considering the two independent trials, except LA 4046, LA 4043, and LA4047, which showed similar behavior. Our findings point out that seeds with low opacity and percentage of damaged seed tissue and high values for living tissue opacity have greater physiological quality. In short, our work confirms the reliability of X-ray imaging and deep learning methodologies in predicting the physiological quality of tomato seeds. timeconsuming, timeconsuming time consuming, consuming time-consuming scale study Xray X ray CNN (CNN MaskRCNN (MaskRCNN categories purpose 4 46 (4 lines seasons 2 analyzed vigorrelated vigor related trials imagebased image based variables 4046 4043 LA4047 behavior short ( 404 LA404 40 LA40 LA4

RESUMO Hancornia speciosa Gomes é uma das espécies frutíferas nativas mais frequentes no Cerrado. Estudos sobre a variabilidade genética de caracteres quantitativos para essa espécie são escassos e a identificação de acessos com as melhores características agronômicas pode subsidiar estratégias de programas de conservação e melhoramento genético. Objetivou-se estimar a diversidade genética de acessos da coleção de germoplasma de H. speciosa da Universidade Federal de Goiás e classificá-los com base em um índice de seleção combinando onze características agronômicas. Foram avaliados 192 acessos individuais de 57 progênies, de 29 populações naturais. Os dados foram submetidos a análise multivariada e os indivíduos e progênies avaliados pelo índice de seleção FAI-BLUP. A análise de agrupamento de Tocher permitiu agrupar as progênies em dez agrupamentos divergentes e os indivíduos em 18 agrupamentos divergentes. A seleção simultânea de caracteres com base no índice FAI-BLUP pode ser recomendada, mantendo-se a variabilidade potencial na população resultante da recombinação. A seleção individual mostrou-se mais eficaz do que a seleção baseada em médias de progênies, por explorar a variação genética entre e dentro das progênies.

ABSTRACT Hancornia speciosa Gomes is one of the native fruit species most frequent in the Brazilian Savanna. Studies on the genetic variability of quantitative traits for this species are scarce and the identification of accessions with the best agronomic traits may support strategies for conservation and breeding programs. This study aimed to estimate the genetic diversity of accessions from the H. speciosa germplasm collection of the Universidade Federal de Goiás (Goiânia, Goiás state, Brazil) and rank them based on a selection index combining eleven agronomic traits. A total of 192 individual accessions from 57 progenies, from 29 natural populations, were evaluated. The data were subjected to multivariate analysis and the individuals and progenies evaluated using the FAI-BLUP selection index. The Tocher cluster analysis allowed grouping the progenies into ten divergent clusters and the individuals into 18 divergent clusters. The simultaneous selection of traits based on the FAI-BLUP index may be recommended while maintaining the potential variability in the population resulting from the recombination. The individual selection proved to be more effective than the selection based on progenies means, because it exploits the genetic variation among and within progenies.

Resumo A produtividade do maracujazeiro no Brasil possui grandes variações entre produtores de diferentes Estados e regiões, em função de tecnologias adotadas. Neste trabalho, objetivou-se avaliar o perfil dos produtores e o uso de tecnologias na atividade produtiva de maracujá, na região do Triângulo Mineiro, impulsionada pela indústria de suco local. A pesquisa foi feita de 2018 a 2019, utilizando a metodologia do Diagnóstico Comportamental da Atividade Produtiva (DCAP), a partir da análise do perfil de 22 produtores, em oito municípios, e os dados foram submetidos à análise estatística descritiva e multivariada. A maioria dos produtores cultiva maracujá em até 5 ha, em regime familiar de trabalho, e possui pouco tempo de experiência. O sistema de cultivo é a céu aberto em espaldeira, com ciclo de produção de 1,5 ano. A maioria dos produtores faz polinização manual, adubação química e orgânica, e não adota, sistematicamente, o manejo integrado de pragas e doenças. As mudas são adquiridas de viveiros habilitados, e a cultivar mais plantada é a FB 300. Os principais prejuízos relatados foram causados por seca e danos no botão floral, com produtividade média de frutos de 17 t ha-1 ano-1. A cadeia produtiva está estruturada, mas há necessidade de intervenções técnicas para aumentar a rentabilidade e a sustentabilidade do sistema de produção.

Abstract Productivity of passion fruit in Brazil varies widely among producers in different states and regions depending on the adoption of technologies. In this study, the goal was to evaluate the producers profile and use of technologies in passion fruit production in the Triângulo Mineiro region, driven by the local juice industry. The research was carried out from 2018 to 2019, using Behavioral Diagnosis of Productive Activity methodology, based on questionnaires applied to 22 producers in eight cities. The data were submitted to descriptive and multivariate statistical analysis. Most producers grow passion fruit in up to 5 ha, in a family work regimen, and have little experience. The cultivation system is open field in vertical trellis, with a production cycle of 1.5 years. The great majority use manual pollination, chemical and organic fertilization, and do not systematically adopt integrated phytosanitary management. The seedlings are purchased from qualified nurseries and the most planted cultivar is the FB 300. The main losses reported by producers were caused by drought and damage to floral bud, with an average fruit yield of 17 t ha-1 year-1. This production chain is structured but requires technical interventions to increase the profitability and sustainability of the production system.

RESUMO Este trabalho teve como objetivo avaliar 15 caracteres agronômicos de 23 genótipos de bananeira dos grupos Cavendish, Prata, Prata-Anã, Pacovan, Mysore e Maçã, em três ciclos de produção, nas condições climáticas da região central de Goiás, Brasil, a fim de identificar os melhores cultivares para essa região. Os tratamentos foram dispostos em delineamento inteiramente casualizado, com repetições variando de 3 a 18 clones de acordo com a disponibilidade de mudas. Por se tratar de dados desbalanceados os componentes de variância foram estimados utilizando a metodologia de modelos mistos. Com o intuito de identificar os genótipos superiores foi levado em consideração os valores genotípicos médios preditos. Os genótipos FHIA-02 (grupo Cavendish), PA42-44 (grupo Prata Anã) e FHIA-18 (grupo Prata Anã), apresentaram maior precocidade de floração. Os genótipos Bucanero, FHIA-17 (grupo Cavendish) e Grande Naine foram os mais produtivos. Os genótipos Grande Naine, FHIA-17 e FHIA-02 apresentam os menores portes e os genótipos FHIA-17, Bucanero e PV79-34 (híbrido de Pacovan) as maiores circunferências do pseudocaule. Os genótipos com as menores taxas de mortalidade foram Pacovan, Garantida (grupo Prata) e Calipso. Com base no índice de seleção, conclui-se que os genótipos FHIA-17, Grande Naine, Bucanero e FHIA-02 foram os mais promissores para serem produzidos na região central de Goiás, por apresentarem alta produtividade e caracteres agronômicos demandados pelos produtores.

ABSTRACT This study had the aim of evaluating 15 agronomic characters of 23 banana genotypes of the Cavendish, Prata, Prata-Anã, Pacovan, Mysore and Maçã groups, in three production cycles, in the climatic conditions of the central region of Goiás state, Brazil, identifying superior cultivars. The treatments were arranged in a completely randomized design, with replicates varying from 3 to 18 clones according to the availability of seedlings. Because of the unbalanced data, the components of variance were estimated using the mixed model methodology. In order to identify the superior genotypes, the predicted mean genotypic values were taken into account. The FHIA-02 (Cavendish group), PA42-44 (Prata Anã group) and FHIA-18 (Prata Anã group) genotypes presented the highest flowering precocity. The Bucanero, FHIA-17 (Cavendish group) and Grande Naine genotypes were the most productive. The Grande Naine, FHIA-17 and FHIA-02 genotypes presented the lowest sizes, and the FHIA-17, Bucanero and PV79-34 (Hybrid Pacovan) genotypes the largest pseudo stems circumferences. The genotypes with the lowest mortality rates were the Pacovan, Garantida (group Prata) and Calipso. Based on the multi-trait selection index, it was concluded that the FHIA-17, Grande Naine, Bucanero, and FHIA-02 genotypes were the most promising cultivars for production in the central region of Goiás due to their agronomic characters and high yield.

Abstract Hancornia speciosa Gomes is a fruit tree native to Brazil with potential for use in cropping systems. The objective of this study was to characterize accessions of a germplasm collection by evaluating agronomic traits. The collection was planted in 2005, in an experiment with 57 progenies and four replications, totaling 192 individual accessions from 29 natural subpopulations of four botanical varieties, from the Brazilian Cerrado. The components of variance were estimated using the restricted maximum likelihood (REML) procedure. Heritability coefficients show the potential of the collection as base population for selection programs. The trait number of fruits per plant presented the highest estimated selection gain. The accessions of H. speciosa var. cuyabensis are the ones with the greatest potential for selection based on plant size and productivity. Progenies with greater development in the juvenile phase are potentially more productive in the adult phase, allowing the early selection of superior genotypes.