The non-linear relation between Evapotranspiration (ET) and the atmospheric moisture demand in terms of vapor pressure deficit (VPD) has been widely reported. Observations point towards a diverse range of potential drivers; however, without uncovering why the identified factors are found to control diurnal ET-VPD hysteresis. Modelling efforts have laid a theoretical foundation to unravel the compound effect of surface and atmospheric states on this hysteresis. However, so far these models have been unable to realistically incorporate the non-linear feedbacks between atmosphere and surface, and lack a vegetation representation that reflects the underlying biological mechanisms. To unravel in what manner the characteristics of ET-VPD hysteresis are controlled, multi-scale observations spanning biology, meteorology, and hydrology from the CloudRoots campaign are combined with modelling in a novel manner. For the first time a proof-of-concept is delivered for reproducing ET-VPD hysteresis with a model that incorporates both surface-atmosphere feedbacks as well as a mechanistic vegetation component. Observations are used for parameter initialization and evaluation to ensure that the modelled hysteresis curve and underlying processes represent a realistic case. Next, this calibrated model is used for a sensitivity analysis. The results show that both the early morning state of the surface and the state of the atmosphere influence the characteristics of the hysteresis loop with respect to its width, height, and initial slope. Via control of the stomatal aperture, soil moisture stress and the vegetation’s capacity to assimilate CO2 for photosynthesis affect the height and width of the curve. The height of the hysteresis loop is significantly affected by entrainment of warm dry air, while its width is minorly impacted. The characteristic fingerprint of entrainment is distinctly different to that of soil moisture stress and the capacity for CO2 uptake. The sensitivity analysis applied to our observationally inspired case, appears to facilitate quantification of the strongest model responses to changing external forcings and parameters. Therefore, the presented approach offers a promising tool, allowing further research into the controlling mechanisms of ET-VPD hysteresis.