Predição por modelos não-lineares do C-CO2 evoluído de argissolo tratado com resíduos orgânicos

Abstract

Many cultural residues and types of management interfere in the decomposition and quality of the soil and crop production. Knowledge of carbon mineralization curves enables us to seek improvements in soil quality and crop productivity. The aim of this study was to compare nonlinear models that describe carbon mineralization and choose the most appropriate, considering surface residue or incorporated into the soil. The data used were obtained from Giacomini et al. (2008) and correspond to the results of an experiment with oat straw, pig slurry and pig deep-litter. The Stanford e Smith, Cabrera and Molina nonlinear models were used, considering autoregressive structure errors AR(1) when necessary. Parameter estimation was performed using the \gnls"statistical tool of software R, using the least squares method and the Gauss-Newton algorithm for convergence. Adjustments were compared using the following evaluators: Akaike Information Criterion (AIC) and Bates and Watts curvature. The Stanford e Smith and Cabrera nonlinear models described satisfactorily carbon mineralization in the soil. The Molina model does not t the data.