The mechanism of water consumption is the basis for management and regulation of water in farmland/orchards. Focusing on the transpiration mechanism of water consumption, the simulation effect of different Jarvis-Stewart model configurations on transpiration consumption in orange forests in dry and hot valleys was compared based on Bayesian parameter estimation methods, and the applicability of Jarvis-〖JP〗Stewart model in the simulation of transpiration water consumption under the condition of strong interaction effect of influence factors was explored. The results showed that considering different influencing factors and their limiting functions would have a great impact on the simulation effect, among which considering soil moisture content and leaf area index had obvious effects on the improvement of the simulation effect, while the introduction of saturated water vapor pressure difference and air temperature would reduce the simulation accuracy to varying degrees. The more impact factors considered, the more complex the model structure was, and the better the simulation effect was. The best model structure screened out basically realized the reliable simulation of water consumption of transpiration in orange forest, but there was still obvious room for improvement in the simulation effect, so the model complexity, simulation accuracy and uncertainty should be comprehensively considered to further explore the appropriate model structure. The research can provide scientific basis for the establishment of water-saving irrigation technology system and water management optimization in orchards, and also provide data support for the further development and improvement of water consumption models.