Abstract: Abstract: Partitioning Evapotranspiration (ET) over an agricultural ecosystem has been one of the most important strategies to improve the water use efficiency of the rice-producing areas in the middle and lower reaches of the Yangtze River in China. However, the traditional ET partitioning presents labor intensive, scales mismatches, and low applicability. The new ET partitioning has been widely expected to consider the water-carbon fluxes coupling in recent years. In the present study, a two-year (2017-2018) water-carbon flux was evaluated using the eddy covariance measurement over a paddy field in the middle and lower reaches of the Yangtze River in China. An Underlying Water Use Efficiency (uWUE) model was also applied to partition the ET into the transpiration (T) and evaporation (E), in order to investigate the seasonal variation of ET and the E/T components. Moreover, the uWUE model was verified by the T/ET observation using the micro-lysimeter and eddy covariance measurement. The results showed that the ET were 318.4 and 335.6 mm for the early rice, and 451.9 and 469.0 mm for the late rice in 2017 and 2018, respectively. The seasonal variation of T/ET was also closely related to the plant phenology. Specifically, the E was the main component of ET with the T/ET ranging from 0.13-0.35 at the regreening stage after the early and late rice were transplanted. The T/ET gradually increased with the growth of the rice plant, when reaching the peak values of 0.62-0.79 at the jointing, booting-heading, and flowering stages. Then, the T/ET decreased with the senescence of the paddy rice. In season, the T/ET values of the early rice were 0.49 and 0.52, respectively, which were lower than those of the late rice (0.62 and 0.55) in 2017 and 2018, respectively, because the light and thermal conditions were much better during the late rice season. Furthermore, the T/ET variations over the paddy fields in the study area were close to those in the Sanjiang Plain, Philippines, Brazil, and Japan, but lower than the dryland ecosystems, such as wheat and maize. In addition, there were great differences in T/ET in the initial period and cloudy days using the uWUE and micro-lysimeter/eddy covariance measurements. The micro-lysimeter/eddy covariance measurement underestimated the T/ET in the early stage, whereas, the uWUE simulation overestimated in cloudy days. However, the daily T/ET presented a significant linear correlation (R2=0.852, P<0.01), indicating that the uWUE model can be used to partition ET over paddy fields. Future work needs to be done to investigate the water use efficiency over paddy fields in humid areas for better ET partitioning. The findings can also provide a strong reference for the application of uWUE model over rice ecosystems, as well as water-saving irrigation and allocation in the middle and lower reaches of the Yangtze River.