Produção do tomateiro: frequências de colheitas e composição de parcela

Abstract

The tomato (Solanum lycopersicum L.) present fractional fruit production commonly described as multiple harvests. The adjustment of the production of the fruit according to the harvests to non-linear models can provide important characteristics for comparison of cultivars. However, the excess of plots with zero production, originated from the use of fixed harvest frequencies, guarantee a high variability of the data. Due to this characteristic, the tomato production databases often do not adhere to the assumptions of homogeneity of variances, independence and Gaussian distribution of the residues, making it difficult to perform statistical analysis based on these assumptions. The composition of the plot by mean of plants, the choice of larger or smaller fixed intervals between harvests and the grouping of the different crop yields may cause influences on this variability. The objective of this work was to evaluate the effect of crop frequency and plot composition on the estimation of two nonlinear models for the adjustment of tomato production and the accumulated tomato production of different cultivars. Tomato production (kg plant-1) of three cultivars (Cordillera, Gaúcho and Janaína) and two plot compositions (average of five plants and seven plants) were measured at two constant crop frequencies (every three days and every six days). The data were obtained by means of experimental field trials and longitudinal planning with measures repeated in time, in the Department of Phytotechnology of the Federal University of Santa Maria, Rio Grande do Sul. Important inference for comparison of cultivars, such as maximum yield and period of this production were estimated by Wood's nonlinear model. Other characteristics such as precocity, production rate and total production were obtained by adjusting the accumulated tomato production to the Logistic growth model. For the estimation of the parameters of these models, the indicated methodology for nonlinear mixed-effect models was used. The heteroskedasticity of the data was adequately reversed by modeling the variance as a function of the dependent variable, and the dependence of the measurements on the observed time for the accumulated tomato production were corrected by the first order autoregressive matrix. Parcels composed of an average of seven plants presented lower variability of the tomato production, independent of the cultivar. The lower variability was also observed for harvests at a higher frequency. The inflection point estimates and the tomato production rate obtained by the parameters of the Logistic model were influenced by the number of plants that compose the plot. In the final production estimate harvest frequency and plot composition did not have significant effects. The maximum production of tomatoes estimated by the Wood model was influenced by the frequency of harvest, and the moment of its occurrence had effect only of the composition of the plot.