Macauba production estimated by regression models

Macauba has been shown to be the most promising native species for oil and biomass production. The first com­mercial crops have arisen mainly in Minas Gerais State, and estimates indicate a vegetable oil yield potential above 8 t ha -1 . In this context, research has acted strongly on technological advance for improvements in the pro­duction system and selection of superior genotypes, which requires the use of precise and less expensive meth­odologies. Thus, we built prediction models to estimate the weight of macauba bunches using easy-to-measure physical variables. Bunches of plants from five regions of Brazil were evaluated. The bunches were weighed, and length and diameter were measured along with four other variables. Based on the set of variables obtained, step­wise multiple regression analysis was used to build regression models for each region. The correlation between observed versus estimated data reached determination coefficients (R 2 ) above 0.90 in three of the models built. The main variables selected by the best models were bunch volume, bunch length/diameter ratio, and bunch square diameter.