Abstract: The Flux Variance Similarity (FVS) method can directly decompose the ecosystem evapotranspiration components based on the observation data of the eddy covariance method. This method needs to calculate the key parameter Ci/Ca (ratio of intercellular CO₂ concentration to atmospheric CO₂ concentration), so the research and construction of the Ci/Ca model plays an important supporting role in the application research of FVS method. In this study, the poplar plantation was taken as the research object, and the transpiration was calculated by the concurrent eddy covariance (CEC) method. Four existing Ci models were inverted, evaluated and optimized: Constant Ci model, Constant Ci/Ca model, Linear model and Katul model. The results showed that the diurnal changes of ecosystem transpiration and soil evaporation based on CEC split showed obvious unimodal curves, and the peak of evaporation appeared before the peak of transpiration. The partitioning results for 83 consecutive days also showed obvious diurnal variation characteristics. TR/ET（transpiration/evaportranspiration）, daily transpiration and daily evaporation are 0.78, 2.46 and 0.63mm, respectively; linear regression deviation was calculated between TR measured by CEC and simulated of FVS method based on Const_Ci model, Const_k model, Linear model and Katul model. The deviation rates were 59%, −16%, −70%, and −31%, MAE were 0.18, 0.069, 0.21 and 0.085mm·h⁻¹, and RMSE were 0.25, 0.089, 0.27 and 0.10mm·h⁻¹ ,respectively. Compared with the original Katul model, the optimized Katul model and VPD0.26 model after parameter optimization have significantly improved accuracy in estimating canopy transpiration. MAE were 0.048 and 0.047mm·h⁻¹, and RMSE were 0.058 and 0.057mm·h⁻¹, and Pearson correlation coefficients were 0.83 and 0.73, respectively. It is feasible to apply the optimized Katul model for FVS to directly separate evapotranspiration components of poplar plantation ecosystem.

Key words: Ci/Ca, Evapotranspiration, Partitioning, Concurrent eddy covariance, Flux variance similarity