Empirical models to estimate the accumulation of dry matter in Marandu palisade grass using agrometeorological variables

Abstract

The objective of this work was to test empirical linear regression models, to predict dry matter accumulation rates (DMAR) of Urochloa brizantha cv. Marandu, using agrometeorological variables. To generate the models, the average dry matter accumulation under rainfed conditions, between 1998 and 2002, was used. The evaluated variables were: minimum, maximum and average temperatures, global radiation (GR), degree-days, actual (AET) and potential evapotranspiration (PET) obtained from the water balance, photothermal units (PU) and the climatic growth index (CGI). Except for the PU, the univariate and multivariate regressions showed good predictive ability. The best results were for the multivariate regression, with T_mín, GR and AET: R², 0.84; root mean square residual (RMSR), 14.72; and Akaike’s information criterium (AIC), 222.5. In the univariate regression, the following variables stood out: corrected degree-days (R², 0.75; RMSR, 17.84; CIA, 242.6), corrected minimum temperature (R², 0.75; RMSR, 17.82; AIC, 244.1); and CGI (R², 0.74; RMSR, 17.85; AIC, 236.9). The correction of the agrometeorological variables using the relation between real and potential evapotranspiration (AET/PET) enhances, in general, the model prediction of DMAR.