Modelling of digesta passage rates in grazing and browsing domestic and wild ruminant herbivores

Abstract

Ruminant utilization of poor-quality feeds is governed by rates of digestion and of passage through the rumen. The passage rate of feed material determines the quantity of bypass nutrients and the efficiency of synthesis of microbial protein in the rumen, making modelling of passage rate important. Artificial neural networks were used to develop models of liquid and solid passage rates. Studies that reported fractional passage rates, along with class and body mass of ruminants, were included in the dataset. Factors that affect rates of passage in all the studies were identified, which included animal and feed factors. The dataset was composed of observations of domestic and wild ruminants of variable body mass (1.5 to 1238 kg) from 74 studies and 17 ruminant species from various climatic regions. Observations were randomly divided into two data subsets: 75% for training and 25% for validation. Developed models accounted for 66 and 82% of the variation in prediction of passage rates for solid and liquid, respectively. On validation with an independent dataset, these models attained 42 and 64% of precision in predicting passage rates for solid and liquid, respectively. Liquid and solid prediction passage rate models had no linear and mean bias in prediction. This study developed better prediction models for solid and liquid passage rates for ruminants fed on a variety of diets and/or feeds from different climatic regions.

Keywords: Artificial neural networks, intake, mean retention time, prediction equation, rumen