Animal Science Department


Date of this Version



J. Anim. Sci. 2015.93:5395–5410; doi:10.2527/jas2015-9220


© 2015 American Society of Animal Science.


The objectives of this study were to evaluate heat stress feed intake models for growing swine using a data set assembled from the literature and to develop a series of new equations modeling the influence of the thermal environment and interactions between the thermal environmental and other factors on feed intake. A literature survey was conducted to identify studies assessing intake responses to temperature. The resulting data set comprised 35 studies containing 120 comparisons to thermoneutral intake. Intake as a fraction of thermoneutral intake (FFI) was the primary response variable, where a value of 1 represented no change from thermoneutral intake. The FFI predicted by NRC and a recent model from a meta-analysis (Renaudeau et al.,) were compared to observed values. New parameters for the NRC equation (NRCmod) were derived, and a series of new equations incorporating duration of exposure (TD), temperature cycling (TC), and floor type (TH) were also derived. Root-mean-square prediction error (RMSPE) and concordance correlation coefficients were used to evaluate all models. The RMSPE for the NRC model was 23.6 with mean and slope bias accounting for 12.6% and 51.1% of prediction error, respectively. The TD, TC, and TH models had reduced RMSPE compared with NRC: 12.9 for TD, 12.6 for TC, and 12.9 for TS. Substantial improvements were also made by refitting parameters (NRCmod; RMSPE 13.0%). In NRCmod, TD, TC, and TH, random error was the predominant source, accounting for over 97% of prediction error. The Renaudeau et al. model was also evaluated. Renaudeau et al. had relatively low RMSPE (22.3) for intake but higher RMSPE for FFI (22.6) than NRC, NRCmod, TD, TC, or TH. Additional parameters were derived for the Renaudeau et al. equation to account for housing system and diet characteristics. This adjustment reduced RMSPE of predicting feed intake (16.0) and FFI (16.3) and reduced systematic bias in the equation. This evaluation of equations highlights the effects of novel explanatory variables on feed intake during heat stress, and the comparison can be useful when selecting a model that best explains variability in feed intake responses to heat stress given available input data.