Sheep productivity in the tropics: finding the limits by a meta-analytic approach

Small ruminants are an important resource for improving the livelihood of smallholder farmers in tropical livestock systems. However, there is a lack of information in regards to the potential of meat productivity (i.e. kg of liveweight produced per area) of growing sheep in such systems, especially when grazing tropical pastures. The aim of this study was to describe the potential of these systems, and identify and quantify the impact of the main factors associated with the two components of meat productivity: average liveweight gain (LWG) per head and number of animals per area. This was achieved by conducting a meta-analysis of published data of post-weaning sheep growth during the wet-season in tropical climates. The empirical data from published studies were collated in a database with the following parameters: stocking rates, grazing method, fertilizer application, grazing time, pasture biomass, pasture species, pasture nutritive value, type of supplementation, level of supplementation, nutritive value of the supplement, animal genotype, sex, initial and final liveweight, liveweight gain and faecal egg count. For this analysis only grazing studies on growing animals which described stocking rate at LWG were selected. The dataset was coded following the recommendations provided by Sauvant et al. (2008) and weighted based on the number of observations. As there was interest in investigating the effects of these management strategies on meat productivity, data were categorised according to the level of nitrogen fertilization and the use or not of supplements. All analyses were performed by specifying a linear mixed effect regression model with study included as a random effect and candidate risk factors included as fixed effects. A backward-step model building process was adopted. The final model that only contained statistically significant main effect terms and, based on Akaike information criterion and conditional and marginal R², considered to best fit the data was selected.