The Thompson model (Thompson et al., in press), a heat balance model for cattle, was evaluated for Bos indicus and B. taurus under different climate conditions through the use of two local and one global sensitivity analyses and tested against independent datasets. The local analyses, which evaluate the individual effects of parameters on model output, showed that the vasodilation/vasoconstriction parameter and reference body temperature (Tbref) strongly affected body temperature. The global analysis, which evaluates the overall effect of parameters on model output, showed that 6 out of 24 parameters account for 0·79–0·89 of the model variation. The high proportion of variation accounted for by the parameters demonstrates that the model is linear in its parameters, with little interaction between the parameters.
The Thompson model was tested against four independent datasets which included both B. indicus and B. taurus animals. The prediction of the relationship between skin and body temperature from the model aligned closely with the relationship in the datasets (R2 ranged from 0·55 to 0·87, mean bias ranged from 0·32 to 1·49). The prediction of sweating and respiration rates from the model aligned closely with the rates measured in the datasets (R2 ranged from 0·80 to 0·98 and 0·79 to 0·93, respectively). The delay in the diurnal body temperature variation, relative to air temperature, was more accurately predicted for cattle in the sun than for cattle in climate chambers. Given the limited datasets for construction and parameterization (both of which are described in Thompson et al., in press), the model evaluated in the current study performed relatively well compared to the literature and known biology.