Pietro André Telatin Paschoalino

The present thesis aims to advance the understanding of changes in Brazilian agricultural diversification in the recent period, at different geographic levels. Furthermore, it intends to assess its relationship with land productivity at the level of geographic microregions. The first chapter brings an introduction about the subject, thus demonstrating the hypotheses, objectives and justifications to carrying out the research. The second chapter assesses the evolution of agricultural crop diversification in Brazil in 2007 and 2017, by measuring the Simpson index and the Effective Number of Species, seeking to identify regions with greater and lesser agricultural diversification (Brazilian microregions) as well as clusters of such regions through Exploratory Spatial Data Analysis. According to the results, there is a decrease in agricultural diversification in Brazil, in addition to a positive spatial autocorrelation in the data. From this finding, it was possible to identify the clusters of regions with high diversification (low) surrounded by regions of also high diversification (low), and the changes in such clusters over time. Finally, despite the two indicators being specific, both showed high similarity, especially when capturing the changes that occurred in the regions in the analyzed periods. The third chapter seeks to analyze the relationship between agricultural diversification and land productivity in 2017 in Brazilian microregions. Thus, the Effective Number of Species was measured as an indicator of agricultural diversification and used as a covariate in the regression on land productivity, as well as other explanatory variables. To assess the correlation between diversification and productivity, the Ordinary Least Squares models and models that consider the spatial dependence of the data were estimated, these being the SAR, SEM and SAC models. From the results, it was noted that diversification had a positive and significant sign on land productivity, and using the Geographically Weighted Regression, a disparity of results (of diversification on land productivity) was identified, depending on the microregion under analysis. In addition, there was a concentration of the positive effects of diversification in the Northeast region, a region recognized for its complexity in relation to its climate. Once verified the positive correlation between agricultural diversification and land productivity in the microregions of the Northeast, in the year 2017, it was sought then, in the subsequent chapter, to specifically evaluate this region. Through panel data regression, which makes it possible to control for the fixed effects of time and individuals, several variables are related to the value of land productivity in the years 2006 and 2017, in which agricultural diversification was used as one of the explanatory variables of the microregion. In addition, as a complementary analysis, it was verified the occurrence of convergence of land productivity in the geographic microregions belonging to that region, between 2006 and 2017, thus using productivity growth in the period as a dependent variable. From the results of the panel data, it was possible to verify a positive relationship of agricultural diversification on land productivity in the region. In addition, it was possible to verify the existence of convergence, and also that agricultural diversification was correlated to productivity growth. In summary, there is, then, a reduction in Brazilian agricultural diversification in the recent period and a positive correlation of diversification on land productivity, with this issue being more robust in the Northeast region, as positive and significant coefficients were found, both through geographically weighted regression and the panel data.