Evaluation of alternative algorithms used to simulate pasture intake in grazing dairy cows

Abstract

Four algorithms used to simulate pasture intake in grazing dairy cows in a dairy decision support system were proposed and evaluated with data from the literature. The algorithms proposed were: 1) an algorithm combining the approach used in a published model to determine dry matter intake based on neutral detergent fiber intake as a percentage of the BW, energy requirements, pasture availability and a standard supplementation (PIest), 2) the previous algorithm modified to consider the type and amount of supplementation (PIsup), 3) an algorithm which considers the effect of selection of pasture (PIsel), and 4) the combination of algorithms 2 and 3 (PIsupsel). Pasture intake data of 27 grazing experiments from the literature were used to evaluate those algorithms. Two methods of evaluation were used: 1) simple linear regression between reported and simulated values, and 2) analysis of variance for the difference between reported and simulated values considering pasture availability and type of supplementation. The R2 of the linear regression and average proportional bias between reported values and simulated values were 0.24 and 19% for PIest, 0.42, and 23% for PIsup, 0.45 and 2% for PIsel and 0.41 and 10% for PIsupsel. Those results showed that PIsel had the lower variability and the values closer to pasture intake. The algorithm PIsup had low variability but tended to underpredict pasture intake. The algorithm PIest values were closer to reported values for low pasture availability. The modeling results show the influence of pasture selection in grazing systems.