ANDRADE, A. S.; http://lattes.cnpq.br/5401409985102127; ANDRADE, Antônia Silânia de.
Resumo:
Soil is a dynamic and complex system to be studied, which requires a considerable number of samples and analyzes for research purposes. With multivariate statistical methods, more favorable conditions can be created by analyzing the samples and the reduction and structural simplification of the data. Having as general objective of the research to evaluate the pattern of spatial distribution of pedological variables (chemical and physical) in the Eastern Cariri of the State of Paraíba. To achieve this, it was first to analyze the chemical and physical attributes of the soil through descriptive statistics techniques and from factor analysis (AF) and thus determine a soil productivity index. Chemical and physical attributes were evaluated, and then the data were subjected to multivariate statistical analysis, also using the Classic Regressive Models (RC); Spatial Regression Auto regressive (MEAR) and Spatial Error (MEE) in order to assess variability and characterize the spatial dependence between some soil attributes in the study area and investigate the association between cotton and bean productivity with agrometeorological variables. When analyzing the attributes after varimax rotation, porosity defines the relationship between the pore volume and the total volume of soil, being influenced by both the texture and the structure of the soil verifying the dendogram, it appears that all groups are differentiated between them based on the joint response of physical and chemical attributes. As for the IPS, it presents a simultaneous and integrated assessment of the soil indicators, recommending that its application consider
fundamentally the pedological classification, with special attention to the soil texture and the climatic classification. Finally, the parameter estimation of the adjusted models was obtained using the Maximum Likelihood method. The performance evaluation of the models was performed based on the coefficient of determination (R2), the maximum value of the likelihood function logarithm and the Schwarz Bayesian information criterion (BIC). This study also allowed to verify the correlation and spatial autocorrelation between the
productivity of cotton, beans and agrometeorological elements, through spatial analysis, using techniques such as Moran's index I. The study was able to demonstrate that, using the performance indicators used, the MEAR and MEE models offered better results in relation to the classic multiple regression model.