MODELOS DE AJUSTE PARA CURVAS DE LACTACIÓN DE VACAS EN CRIANZA INTENSIVA EN LA CUENCA DE LIMA

Abstract

The description of lactation curves by mathematical models could be an effective tool for management decisions in dairy cattle production systems. In this study the comparative fit of three models: negative exponential (í = (â0e-â1x), incomplete gamma (í = â0xâ1e-â2x) and fifth order polynomial (í = â0 + â1x + ... +â5x5) was evaluated both between and within lactations using 8,763 milk yield records from 770 lactations for the first two models, and 7,228 records from 647 lactations for the third model. Milk yield records included the period 1997-2001 and were collected from five dairy farms located in Lima. The models were compared on the basis of the adjusted multiple coefficient of determination (R2 aj), the residual standard errors (EER), the standard errors of the predicted yield on day 305 (EE305), and the residual scatter plots. Additionally, the effect of parity, calving season and the 1998 El Niño phenomenon (FEN) on the basic and derived parameters of the models were estimated using general lineal models. In terms of fitness, the models ranked fifth order polynomial > incomplete gamma > negative exponential both between and within lactations. However, the fifth order polynomial model had several shortcomings like higher EE305, a tendency of overestimate late lactation yields, and a requirement of a minimum number of records. These left incomplete gamma as the model of choice, being capable to explain 72% of the yield variation within lactations and having an EE305 of only 3.3 kg. Parity and calving season affected significantly most of the basic and derived parameters of the models, whereas the FEN only affected the b0 parameter of the exponential negative and the peak yield predicted by the incomplete gamma model. All of the three mentioned factors had significant effect on the total yield predicted for a 305 day-lactation.