Abstract
A meta-analysis was conducted to elucidate the factors affecting drinking water intake (DWI) in order to develop new DWI empricial predictive models for lactating dairy cows and to compare with existing models. A large dataset, containing 262 experimental diets from 75 experiments published in 65 peer-reviewed papers was built. Selected explanatory variables were grouped into animal (body weight (BW); BW0.75; milk yield), diet composition (dry matter; ash; sodium; potassium; crude protein (CP); neutral detergent fiber), dry matter intake (DMI) and environment (mean air temperature and relative humidity) inputs. To develop the predictive models, the dataset (peer-reviewed papers) was randomly divided into two subsets for statistical analyses. The first data subset was used to develop equations to predict DWI (41 peer-reviewed papers; 47 experiments; 157 experimental diets), and the second data subset was used to assess the adequacy of the predictive models (24 peer-reviewed papers; 28 experiments; 105 experimental diets). Ash was the main diet input affecting DWI in lactating dairy cows, while BW0.75 affected DWI more than milk yield. Use of dietary characteristics (DM, ash and CP) and DMI as inputs could be enough to predict DWI as accurately as more complete models. Dry matter intake as predictive variable does not improve DWI prediction in models with only animal (BW0.75 and milk yield) as input, but it substantially improves prediction of models without use of animal as input. Among the existing models evaluated, those developed from just one experiment (Castle and Thomas, 1975; Little and Shaw, 1978; Murphy et al., 1983; Stockdale and King, 1983; Dahlborn et al., 1998) had poor prediction quality. We proposed the use of two models (diet II and animal + diet models) that presented the lowest root mean square prediction error (RMSPE) to predict DWI in lactating dairy cows.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have