Why Plant Hydraulics is Necessary to Predict Evapotranspiration Under Soil Water Stress

Abstract:

Understanding plant water use, or evapotranspiration (ET), response to soil water stress is crucial for accurately predicting the energy, water, and carbon cycles across spatial or temporal scales. Until recently, many hydrologic, dynamic vegetation, and land surface models used a single empirical correction function (β), parametrized to a specific soil-plant system, to downregulate ET in response to soil water stress. Despite its parsimony, there is mounting evidence that β leads to large prediction uncertainties, which has led many in the modeling community to turn to more physically based plant hydraulics models (PHMs). PHMs more accurately represent transport through the soil-plant-atmosphere continuum by coupling soil water supply to atmospheric moisture demand through their mutual dependence on leaf water potential. However, with the increased realism comes additional challenges associated with parameter uncertainties and computational complexity. Thus, the relative strengths and weaknesses of PHMs and β motivate a better understanding of their theoretical foundations as well as the conditions under which PHMs' complexity and realism is justified over the simplicity of β.

We analyze a PHM and β using a supply-demand framework and demonstrate that: i) ET lies on a spectrum of hydraulic transport limitation, ii) β represents the end-member scenario of PHMs where there is negligible resistance to flow in the soil-plant system, and iii) in transport-limited systems, atmospheric moisture demand and soil moisture further enhance the differences between the two models. By comparing predictions from a calibrated land surface model at an Ameriflux site using PHM versus β downregulation schemes, we further demonstrate that the errors incurred by β in lieu of PHMs at this transport-limited site are concentrated during times of the year with low water supply and amplified by high atmospheric moisture demand. However, we show these errors can be mitigated by using a dynamic β function that has additional dependence on atmospheric demand, creating a compromise between computational realism and parsimony. This work highlights the limitation of the β scheme under dynamic soil and atmospheric conditions and the importance of utilizing PHMs or a dynamic β for sites subject to hydraulic transport limitation.