标题：Estimativa da evapotranspirao de referência do Planalto Paulista por meio de regresses múltiplas com dados ausentes estimados via análise de componentes principais
出版社：Grupo de Pesquisa Metodologias em Ensino e Aprendizagem em Ciências
摘要：Evapotranspiration is a physical phenomenon that promotes the complex transfer of water to the atmosphere through the relationship between climatological water balance, surface water evaporation and transpiration from agricultural crops. Obtaining reliable measurements of evapotranspiration is a complex task. The objective of this work was to apply the multivariate technique Principal Component Analysis to fill in missing data and model the reference evapotranspiration by multiple regression, with subsequent comparison with the Penman-Monteith model. Principal Component Analysis, the EM algorithm, was applied to reconstruct the climatic database of 30 automatic weather stations in the Western Plateau of So Paulo, located in the northwest of the State of So Paulo, Brazil. Measures for the period 2013-2017. Subsequently, an exploratory analysis of the climatic variables was carried out to verify the grouping of the most relevant climatic variables in the physical processes of evapotranspiration. These clusters were the basis for the construction of different reference evapotranspiration estimation models through Multiple Regressions. The results showed the best performances for EToRLM4 (rRMSE = 5.23%), EToRNLM4 (rRMSE = 6.39%). The values of the statistical indicatives of the validation base of RLM4 and RNLM4 indicate that both multiple regression models can be used to estimate the reference evapotranspiration.
