Variabilidade espacial da produtividade da soja na região oeste do Paraná associada a variáveis agrometeorológicas utilizando Bootstrap

• Ano de defesa: 2020
• Autor(a) principal: Gabriel, Caroline Cristina Engel
• Orientador(a): Uribe Opazo, Miguel Angel
• Banca de defesa: Uribe Opazo, Miguel Angel , Araújo, Everton Coimbra de , Cima, Elizabeth Giron , Oliveira, Marcio Paulo de , Christ, Divair
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual do Oeste do Paraná; Cascavel
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Agrícola
• Departamento: Centro de Ciências Exatas e Tecnológicas
• País: Brasil
• Palavras-chave em Português: Bootstrap; Índice Ι de Moran; Modelo SAR; Modelo CAR
• Palavras-chave em Inglês: Bootstrap; CAR Model; SAR Model; Moran Global I index
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::ENGENHARIA AGRICOLA
• Link de acesso: http://tede.unioeste.br/handle/tede/5029

Monitoring soybean productivity is essential for market strategies and world food security. Considering crop forecasts, it is paramount to understand the relationship among factors that influence the growth of the crop from a space-time perspective, with closer attention to the agrometeorological aspects which affect the soy cycle. Spatial area statistics search for patterns of association in a given location through spatial correlation indices. Once the correlation of factors in soybean productivity is detected, it is possible to make predictions using spatial regression models, which incorporate the spatial structure of the data into the model. However, in small samples it is difficult to meet the assumptions of these models; thus, exploring the bootstrap resampling technique of the residues as an estimator of the coefficients is a valid option. Therefore, a spatial analysis of soybean productivity and the agrometeorological variables, mean temperature, mean global solar radiation, and pluvial precipitation were performed using the Moran Global I indexes and the Local Spatial Association Index (LISA), to predict soybean productivity through the Spatial AutoRegressive (SAR) and Conditional AutoRegressive (CAR) space models in the 2014/2015, 2015/2016 and 2016/2017 crop years, in the Western region of Paraná. It was observed that soybean productivity is strongly autocorrelated in the cities of the study, and the significance of the correlation indexes confirmed the influence of agrometeorological variables on soybean productivity. In addition, when using the agrometeorological variables in the productivity models, the best values were found for the Akaike Information Criteria – AIC, the Maximum Value of the Logarithm of the Likelihood Function – MVLFV, and the Root Mean Square Error – RMSE with the CAR model in the 2015/2016 crop year. The analysis of the SAR and CAR models from the estimates by the bootstrap technique resulted in values of the coefficients approximate to those estimated by maximum likelihood, which corroborates the use of the bootstrap of the residues.