Evapotranspiration partitioning for three agro-ecosystems with contrasting moisture conditions: a comparison of an isotope method and a two-source model calculation

Quantification of the contribution of transpiration (T) to evapotranspiration (ET) is a requirement for understanding changes in carbon assimilation and water cycling in a changing environment. So far, few studies have examined seasonal variability of T/ET and compared different ET partitioning methods under natural conditions across diverse agro-ecosystems. In this study, we apply a two-source model to partition ET for three agro-ecosystems (rice, wheat and corn). The model is coupled with a plant physiology scheme for the canopy conductance. The model-estimated T/ET ranges from 0 to 1, with a near continuous increase over time in the early growing season when leaf area index (LAI) is less than 2.5 and then convergence towards a stable value beyond LAI of 2.5. The seasonal change in T/ET can be described well as a function of LAI, implying that LAI is a first-order factor affecting ET partitioning. Application of the model to seven other Ameriflux sites reveals that soil moisture and canopy conductance also influence the ET partitioning. The two-source model results show that the growing-season (May–September for rice, April–June for wheat and June–September for corn) T/ET is 0.50, 0.84 and 0.64, while an isotopic approach shows that T/ET is 0.74, 0.93 and 0.81 for rice, wheat and maize, respectively. The two-source model results are supported by soil lysimeter and eddy covariance measurements made during the same time period for wheat (0.87). Uncertainty analysis suggests that further improvements to the Craig-Gordon model prediction of the evaporation isotope composition and to measurement of the isotopic composition of ET are necessary to achieve accurate flux partitioning at the ecosystem scale using water isotopes as tracers.The program code for the two-source model is available at the open-source platform https://github.com/zhongwangwei/SiLSM_v3.